The volume-outcome relationship and minimum volume standards--empirical evidence for Germany.

HEALTH ECONOMICS Health Econ. 24: 644–658 (2015) Published online 3 April 2014 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/hec.3051

THE VOLUME-OUTCOME RELATIONSHIP AND MINIMUM VOLUME STANDARDS – EMPIRICAL EVIDENCE FOR GERMANY CORINNA HENTSCHKERa,b and ROMAN MENNICKENa, a Rheinisch-Westfälisches

Institut für Wirtschaftsforschung, Essen, Germany

b Ruhr-University

Bochum, Germany

SUMMARY For decades, there is an ongoing discussion about the quality of hospital care leading i.a. to the introduction of minimum volume standards in various countries. In this paper, we analyze the volume-outcome relationship for patients with intact abdominal aortic aneurysm and hip fracture. We define hypothetical minimum volume standards in both conditions and assess consequences for access to hospital services in Germany. The results show clearly that patients treated in hospitals with a higher case volume have on average a significant lower probability of death in both conditions. Furthermore, we show that the hypothetical minimum volume standards do not compromise overall access measured with changes in travel times. Copyright © 2014 John Wiley & Sons, Ltd. Received 22 April 2013; Revised 19 February 2014; Accepted 25 February 2014 KEY WORDS:

volume-outcome; minimum volume standards; hospital quality; mortality; Germany

1. INTRODUCTION For decades, there is an ongoing discussion about the quality of hospital care. The publication of To Err is Human: Building a Safer Health System by the Institute of Medicine (2000) was followed by major efforts to improve hospital safety, leading inter alia to the introduction or recommendation of minimum volume standards for specific elective conditions in the USA (Leapfrog Group, 2011), Scotland (Healthcare Improvement Scotland, 2011), or Germany (G-BA, 2013). Minimum volume standards ground on the volume-outcome theory, which hypothesizes that hospitals with a higher case volume obtain a better outcome quality, that is, the manufacturing’s learning curve is applied to service organizations (Luft et al., 1987). Accumulated case volume of a hospital leads to decreasing adverse event rates through improvement of skills, standardization, and better organization (Birkmeyer et al., 2003a; Gandjour and Lauterbach, 2003). Later, if learning reaches its plateau, high-volume hospitals are able to perform a procedure more regularly and therefore maintain their high level of learning (Gandjour and Lauterbach, 2003). Furthermore, economies of scale may play an important role in this context. Larger hospitals might have a broader skill-mix in staffing and a better training environment, which could positively impact health outcomes. Additionally, hospitals with higher volumes can more easily afford better technical equipment. These concepts are summarized under the ‘practice-makes-perfect’ hypothesis (Gaynor et al., 2005). The volume-outcome relationship has been controversially discussed in the literature. For the first time, Luft et al. (1979) examined this relationship empirically and found an inverse relation between mortality and case volume in 10 out of 12 conditions. Numerous further publications followed with a focus on outcome

Correspondence to: RWI, Hohenzollernstr. 1-3, 45128 Essen, Germany. E-mail: [email protected]

Copyright © 2014 John Wiley & Sons, Ltd.

VOLUME-OUTCOME AND MINIMUM VOLUME STANDARDS

645

quality mostly measured as in-hospital mortality1 (Hughes et al., 1988; Wen et al., 1996; Birkmeyer et al., 2002; Barker et al., 2011), but also regarding other consequences of regionalization. Ho et al. (2007) and Ho et al. (2012), for example, conclude that the positive effect of a higher volume on social welfare is attenuated by increasing prices due to reduced competition in the market.2 The majority of analyses focuses on specific conditions ranging from surgical or orthopedic interventions for patients with heart diseases (Thiemann et al., 1999), cancer (Patti et al., 1998), or hip and knee fractures (Taylor et al., 1997) to the treatment of patients with AIDS (Hogg et al., 1998). In this study, we focus on two important conditions, abdominal aortic aneurysm without rupture (AAA) and hip fracture (HIP). HIP3 is one of the most frequent hospitalization reasons among older age groups. Because of the high incidence more than 1,200 hospitals in Germany treat those patients. In contrast, AAA4 occurs far less frequently, that is, treatment is performed by less than 400 hospitals. However, the treatment of AAA is complex, costly, and every mistake in the treatment process can lead to clinical complications (AHRQ, 2007). Hence, the two indications differ in prevalence and therefore volume rates as well as the number of hospitals in which the conditions are treated. Another distinguishing factor of the two diagnoses is the urgency: AAA without rupture are basically elective interventions. HIPs are emergencies with a ‘deferrable priority’ (Rommens, 2001). Most empirical studies that examine the volume-outcome relationship of intact AAA found that the case volume of a hospital has a significant effect on quality (e.g., Birkmeyer et al., 2002; Dimick and Upchurch, 2008; Holt et al., 2009; Wen et al., 1996). Recent publications distinguish between patients with an open and an endovascular aneurysm repair (EVAR) (Dimick and Upchurch, 2008; Holt et al., 2009; Landon et al., 2010; McPhee et al., 2011). So far, evidence from outside the USA is sparse. For example, we could only identify one study by Eckstein et al. (2007) using German data. Hip fracture is less frequently evaluated in volume-outcome studies, most likely because studies usually find no significant relationship between volume and outcome (Browne et al., 2009; Hamilton and Hamilton, 1997; Hamilton and Ho, 1998; Sund, 2010; Wenning et al., 2000). Only Hughes et al. (1988) and Forte et al. (2010) find a significant volume-outcome effect for patients with a HIP; however, the study by Hughes et al. has been criticized for insufficient risk adjustment (Hamilton and Hamilton, 1997). The volume-outcome relationship is the rationale for minimum volume standards in hospitals. One concern against minimum volume standards is that they could endanger access to hospital services. At the moment, two thirds of all German residents can reach the next hospital within 10 min, and 97.5% within 20 min. For merely 2.5% of the residents, it takes more than 20 min to reach the closest hospital from their residence (BBSR, 2011). However, these calculations neglect the fact that not all hospitals provide services for all conditions, that is, especially elective patients cannot just go to the next hospital available if it does not offer treatment for their specific condition. Hence, assessing access to hospital services under the assumption that all hospitals provide care for all patients will grossly underestimate ‘actual’ travel times for specific conditions. Minimum volume standards have the consequence that some hospitals do not provide a special service anymore causing increasing travel times for patients. For example, Geraedts et al. (2008) showed that with the introduction of a minimum volume in esophagus interventions the average distance increased from 34 km in 2004 to 42 km in 2006. Given that these results are based on extrapolated data and are unweighted by number of inhabitants, the authors refrained from deriving any ‘scientifically founded conclusions about the appropriateness of minimum volume standards’ (p. 895). The German minimum volume regulation states that a nationwide supply with hospital 1

Further quality indicators are complications rates, length of stay, or death within a defined time interval (Browne et al., 2009; Forte et al., 2010). However, in Germany, prices for hospital services are fixed by state regulations; that is, providers are not allowed to change their prices for services. Hence, we do not consider the effects of reductions in hospital competition. 3 Hip fractures are fractures in the area of the hip, which are mostly caused by falls. Treatment includes either a conservative therapy, a joint preserving therapy, or a total or half replacement of the hip joint (Beck and Rüter, 1998). 4 Abdominal aortic aneurysms are dilatations of the abdominal aorta of over 3 cm. An operation is necessary if the dilatation exceeds 5 cm. In this case, treatment is provided either with an open operation or with an endovascular aneurysm repair to avoid a rupture of the abdominal aorta (Torsello et al., 2005). 2


Health Econ. 24: 644–658 (2015) DOI: 10.1002/hec

646

C. HENTSCHKER AND R. MENNICKEN

services has to be guaranteed, but so far, there is no definition for a travel time, which must not be exceeded. Only a few federal states have defined travel times for emergencies. For example in Hesse, emergencies are supposed to be in a hospital within 20 min but not more than 30 min (Hessisches Sozialministerium, 2008). There is no definition of travel times regarding elective treatments. With this study, we would like to contribute to the international body of evidence, which is mainly focussed on USA. We are the first to present a large scale study using a full sample of all patients with intact AAA and HIP using German data. The data has the advantage that we have a full in-patient sample available, that is, analyses are not restricted to a specific patient group such as, for example, only Medicare patients. We provide new evidence for a volume-outcome relationship for patients with a HIP and support the existing evidence for AAA of a significant correlation between volume and outcome. Furthermore, we define hypothetical minimum volume standards for both conditions and assess changes in access to hospital services. We compare average travel times before and after the introduction of the minimum volume standard. The remaining part of this paper is organized as follows. Section 2 describes the data, the estimation strategy, and the minimum volume standards. Results are shown in Section 3. Section 4 discusses the results and concludes. 2. DATA AND METHODS 2.1. Data The analyses are based on administrative data coming from the German system of diagnosis related groups (DRGs) of about 18.6 million hospital cases of 1780 German hospitals for the year 2007. The data comprises all in-patients in Germany except psychiatric cases and is originally collected for billing purposes toward health insurance companies. They include detailed information on patient characteristics such as age, gender, ZIP code, length of stay with admission and discharge date and status, main diagnosis, and secondary diagnoses given with the respective German ICD-10 codes. Additionally, the data contain information on hospital level, specifically address, ownership type (private not-for-profit, private for-profit, and public), bed capacity, and teaching status. We use in-hospital mortality as a quality indicator. For both conditions, mortality is one of the in-patient quality indicators approved by the Agency for Healthcare Research and Quality (AHRQ, 2007) and can therefore be used to examine differences in quality between hospitals. Mortality is a common quality indicator (Birkmeyer et al., 2002; Keeler et al., 1992) as it is the most serious clinical outcome. Additionally, in contrast to other outcome variables, mortality is regarded as robust against different coding behavior of hospitals (AOKBundesverband et al., 2007). The latter point applies to the quality of the data set. Every hospital records its own data. As hospitals know that the data are used for a qualitative evaluation, it is possible that adverse events, for example, complications, are not coded (Romano et al., 2002). Thus, we would not only evaluate quality differences but also coding differences between hospitals. Unfortunately, we cannot track patients after their hospital stay as we have only hospital data available. Hence, it is not possible to use out-of-hospital mortality as quality indicator, and our results could be biased by the discharge practice of hospitals. Thereby, we assume that hospitals do not discharge patients who are shortly before dying. We think that even if the mortality rate after discharge is increasing, a potential volume-outcome effect remains (Forte et al., 2010; Rigberg et al., 2006). We include patients with main diagnosis5 of AAA or HIP and a matching procedure code. For identifying patients with AAA, we use diagnosis and procedure codes, which were defined by the German Federal Joint Committee in their quality assurance agreement for this condition (G-BA, 2010). This includes patients with an intact AAA who obtain an open or EVAR. Patients with ruptured AAA were excluded from the analysis (e.g., Dimick and Upchurch 2008; Wen et al., 1996). We identified diagnoses for HIP patients

5

The German diagnosis related groups system defines the main diagnosis as ‘the diagnosis that is identified as primarily responsible for causing the hospital stay of the patient’ (DKG et al., 2006).




647

by using the definition of the Federal Office for Quality Assurance (BQS, 2008) and Browne et al. (2009). HIP patients have either a femoral neck fracture or a pertrochanteric fracture who undergo an open or closed reposition or obtain an implantation of an endoprothesis. Patients with a matching procedure code but AAA or HIP only coded as a secondary diagnosis are not included. These patients differ significantly from patients with the condition as main diagnosis. They have, for example, a higher mortality rate and a longer length of stay. Hence, it has to be assumed that the outcome of these patients is primarily determined by their main diagnosis. We only include patients who are at least 20 years old, because younger patients might need a special or different treatment. Patients who are transferred to another hospital are excluded (AAA, n D 245; HIP, n D 9210). For these patients, we cannot determine the quality of the transferring hospital because we do not know the reason for the dismissal. Furthermore, we exclude patients who are treated in hospitals with less than 3 AAA patients (n D 136) or less than 10 HIP patients (n D 338) in 2007. Estimations with full sample did not qualitatively change our results. Accordingly, our final sample consists of 7980 AAA patients from 386 hospitals and of 89203 HIP patients from 1155 hospitals. We classify the case volume in quintiles.6 This approach ensures that a sufficient case volume exists in every quintile. As a result, we can distinguish between patients who are treated in hospitals with a very low, low, medium, high, and very high case volume. Except for volume, the treatment effect is furthermore influenced by risk factors of the patients. For this reason, we control for age and gender. Additionally, we include indicator variables for admissions during winter time, for weekend or holiday admissions, and for patients who were transferred between departments during the hospital stay. The first variable should capture possible seasonal patterns especially for HIP patients. It is conceivable that because of severer falls during winter time, more complications arise. The second variable captures possible weekend or holiday effects due to a lower staffing level on these days (Bell and Redelmeier, 2001). Transfers during a hospital stay might also increase the risk of death independent of the case volume, for example, due to a loss of information between departments. Furthermore, we include an indicator variable for the type of HIP (only for HIP patients) and for the type of operative procedure (only for AAA patients). We further control for the admission status (scheduled, emergency, and transfer). To account for number and severity of secondary diagnoses, we use the Charlson comorbidity index (CCI) (Charlson et al., 1987). The higher the CCI is, the more ill the patient is besides having AAA or HIP. This is a standard approach for risk adjustment in the literature to account for observable patient heterogeneity. For this purpose, we use diagnosis codes of Quan et al. (2005) who mapped the original codes from ICD-9 system into the ICD-10 system. Another possibility for comorbidity risk adjustment is the method of Elixhauser et al. (1998). Both models lead to similar results. Therefore, only estimates with CCI are presented in this paper. Other studies have shown that besides volume, further hospital characteristics, for example, ownership (Milcent, 2005), teaching status (Ayanian and Weissman, 2002), location, and size of the hospital (Keeler et al., 1992), could also affect quality of care. Hence, we include the mentioned variables. We further add a variable controlling for an existing intensive care unit to serve as a proxy for technical equipment of the respective hospital. All variables are explained in Table A1 in the Appendix. Tables I and II show descriptive statistics of variables used in the analysis for AAA and HIP, respectively. The overall unadjusted in-hospital mortality rate is 3.5% for AAA repair and 6.3% for HIP treatment, which are high mortality rates compared with an average hospital stay. The yearly case volume for the treatment of AAA ranges from 3 to 209 and for HIP from 10 to 387. Because of the classification of case volume in quintiles, we end up with an unequal distribution of the number of hospitals. In the quintile with the lowest case volume, 54.1% of the hospitals treat 20.8% of the patients with AAA. In contrast, 18.8% of the patients with AAA are treated in only 3.9% of the hospitals with the highest case volume. This is similar for the distribution of hospitals with HIP treatment.

6

This is a classification that is often used in the volume-outcome literature. Nevertheless, this is an arbitrary division. Therefore, we also specified the case volume as a continuous variable and in tertiles. The case volume effect is robust for all different classifications.



648


Table I. Descriptive statistics of AAA Variable (Number of cases) Patients (n) Deaths (%) Hospitals (n) Hospitals (%) EVAR (%) Male (%) Age (years) Admission reason (%) Scheduled admission Emergency Transfer Charlson comorbidity index (%) 0 1–2 3–4 5 Transfer between departments (%) Winter (%) Weekend or holiday admission (%) Ownership (%) Public Private non-profit Private for-profit Teaching hospital (%) University hospital (%) Beds (%) 200 beds 201–500 beds > 500 beds ICU (%) Urban (%)

Total

1 (3–15)

Case volume quintile (n = 7980) 2 3 4 (16–25) (26–39) (40–67)

5 (68–209)

7980 3.5 386 100.0 39.7 89.3 70.8

1656 5.2 209 54.1 32.1 88.8 70.8

1562 4.4 78 20.2 37.4 89.1 71.1

1580 2.8 52 13.5 39.6 88.5 70.5

1683 2.6 32 8.3 38.6 90.2 70.6

1499 2.5 15 3.9 51.8 90.0 70.9

83.7 12.6 3.7

83.8 13.9 2.4

83.4 13.7 2.9

84.8 13.0 2.2

88.3 8.0 3.7

77.7 14.9 7.4

21.1 50.8 21.2 6.9 47.6 31.9 8.0

21.5 49.3 22.1 7.1 52.5 32.7 9.9

23.4 49.4 20.5 6.7 47.4 32.4 6.6

19.4 50.8 22.3 7.5 55.4 32.3 5.8

17.2 53.7 21.3 7.8 46.6 29.3 7.9

24.4 50.9 19.7 4.9 35.2 33.0 9.7

50.4 31.9 17.7 75.6 16.0

42.9 40.4 16.7 67.7 4.0

50.2 39.4 10.4 72.4 11.7

51.4 34.6 14.0 82.7 13.4

54.1 26.4 19.4 81.4 29.1

53.8 18.0 28.2 73.6 21.7

3.9 33.1 62.9 55.1 62.9

6.5 52.8 40.8 49.3 65.8

7.6 37.4 55.0 48.2 85.1

1.7 29.2 69.1 53.5 92.5

3.6 28.7 67.7 55.8 93.2

0.0 16.2 83.8 69.7 95.5

AAA, abdominal aortic aneurysm; EVAR, endovascular aneurysm repair; ICU, intensive care unit.

Four thousand, eight hundred thirteen patients with AAA (60.3%) underwent an open aneurysm repair and 3167 patients (39.7%) underwent EVAR. Though, the proportion of EVAR is in hospitals with a very high case volume with 51.8% the highest. On average patients with AAA are 71 years old, are male, and have a CCI of 1.8. Basically, AAA is an elective intervention with only 12.6% of the patients admitted as emergencies. Patients with HIP are on average 80 years old, are female, and have a CCI of 1.4, and 76 % are emergency admission. In general, patients treated in hospitals with a higher case volume are more often treated in larger and urban hospitals. 2.2. Econometric specification For the empirical analysis, we use multiple logistic regression analysis with the dependent variable yih indicating whether patient i has died in hospital h (yih D 1) or not (yih D 0). We estimate the following equation via maximum likelihood estimation: yih D ˛0 C vol 0 h ˇ 1 C x 0 ih ˇ 2 C k0 h ˇ 3 C "ih (1) yih D 1 if yih 0; where yih is an unobserved latent variable, which determines whether the patient died in hospital, vol is a categorical variable for the case volume, x are the patient characteristics, k are the hospital characteristics, and " is a standard logistically distributed error term.



649


Table II. Descriptive statistics of HIP Variable (Number of cases) Patients (n) Deaths (%) Hospitals (n) Hospitals (%) Femoral neck fracture (%) Male (%) Age (years) Admission reason (%) Scheduled admission Emergency Transfer Charlson comorbidity index (%) 0 1–2 3–4 5 Transfer between departments (%) Winter (%) Weekend or holiday admission (%) Ownership (%) Public Private non-profit Private for-profit Teaching hospital (%) University hospital (%) Beds (%) 200 beds 201–500 beds > 500 beds ICU (%) Urban (%)

Total

1 (10–58)

Case volume quintile (n = 89,203) 2 3 4 (59–81) (82–109) (110–150)

5 (151–387)

89203 6.3 1155 100.0 53.7 25.1 79.6

18468 7.0 484 41.9 56.0 25.1 79.4

17281 6.8 247 21.4 54.6 24.2 79.6

18144 6.7 192 16.6 53.6 25.2 79.4

17587 5.7 139 12.0 52.1 25.7 79.5

17723 5.3 93 8.1 52.1 25.2 79.9

21.8 76.2 2.0

26.3 71.7 2.0

23.2 75.3 1.6

21.9 75.8 2.3

18.2 79.8 2.0

19.2 78.6 2.2

35.7 43.3 15.3 5.7 27.1 33.1 28.4

35.7 43.4 15.3 5.6 20.5 33.3 27.7

35.3 44.3 14.4 5.9 24.3 33.2 28.4

36.1 43.2 15.0 5.8 26.4 33.2 28.8

36.0 43.0 15.5 5.5 28.0 32.7 28.7

35.3 42.9 16.1 5.7 36.6 33.4 28.7

49.4 37.6 13.1 55.6 3.8

36.4 48.3 15.4 22.1 2.2

42.1 47.3 10.7 37.7 2.5

48.5 37.3 14.3 55.5 4.6

55.2 32.4 12.4 76.7 3.6

65.2 22.4 12.3 87.1 6.0

19.7 47.8 32.5 44.8 32.5

53.4 39.7 6.9 26.2 59.2

29.0 56.0 15.0 34.1 72.1

9.4 61.3 29.3 43.5 72.6

2.7 50.5 46.8 54.9 76.5

3.0 31.8 65.2 65.9 88.5

HIP, hip fracture; ICU, intensive care unit.

We use cluster-robust standard errors to account for the dependency of patients in the same hospital. Another possibility to account for the correlation of the error terms of patients in the same hospital is to add a random intercept to the model. Both approaches lead to similar results. Therefore, we only present results with clusterrobust standard errors. However, it is important to account for the dependency of patients in the same hospital. Otherwise, it leads to an underestimation of the variance or standard errors as the dependency of patients in the same hospitals is ignored. An issue often neglected as in, for example, Browne et al. (2009), McPhee et al. (2011), Wen et al. (1996), or Wenning et al. (2000). We measure model performance with the c-statistic. The c-statistic specifies how well a model can differentiate between patients with different outcomes. Values above 0.7 indicate an acceptable discrimination, and values above 0.8 indicate an excellent discrimination (Hosmer and Lemeshow, 2000). 2.3. Minimum volume standards We supplement our analysis by showing changes in access to hospitals if a minimum volume standard is introduced. Addresses of hospitals and the centroids of all ZIP codes were geo-coded. Geographic centroids correspond with the geographic center of each ZIP code area. We assume that all patients within a specific ZIP code area live at the geographic centroid and patients’ ZIP codes were based on their home address. We use this information to calculate travel times for each patient to both, the hospital chosen for treatment and all further Copyright © 2014 John Wiley & Sons, Ltd.


650


hospitals around the patient’s ZIP code area.7 We calculate driving time by car in minutes and street kilometers, that is, we take geographic and infrastructural differences into account. This point is often neglected but is especially important for more rural areas because straight line measurements would underestimate access in regions with less comprehensive infrastructure. We present all results using driving time by car in minutes. If a hospital has a smaller case volume than the hypothetical minimum volume standard, this hospital loses authorization to treat these patients, that is, it is closed for services, and all patients treated there seek then treatment in the hospital nearest to their ZIP code still ‘open’ for treatment. As patients are fairly free to chose a hospital for treatment in Germany, a significant number of patients do not seek treatment in the nearest hospital from their residency. Hence, diverting patients from a ‘closed’ hospital to the nearest one still ‘open’ could lead to a decrease in average travel times. However, we are interested in potential changes in access by minimum volume standards, so we follow Kansagra et al. (2004) by setting all patients’ travel times to the minimum. Accordingly, all patients are treated as if they had chosen the nearest hospitals for treatment, that is, we change travel times to the minimum possible for 33897 (or 38%) of HIP and for 3431 (or 57%) of AAA patients. All other, that is, 62% of HIP and 43% of AAA patients have chosen the nearest hospital for treatment. This approach ensures that travel times increase under a minimum volume standard. Nevertheless, when showing changes in travel times due to hypothetical minimum volume standards, we also include ‘actual’, that is, unchanged travel times for comparison. We set hypothetical minimum standards to a minimum of 15 AAA and 58 HIP patients for each hospital, that is, all hospitals in the first case volume quintile of each condition exit the market immediately (scenario 1).8 Furthermore, we assume a stepwise introduction of minimum volume standards to allow hospitals to adjust (scenario 2). We do not divert all patients from the hospitals below the minimum volume standard immediately, but only the patients from hospitals in the lowest quartile below the hypothetical standard. Thus, all remaining hospitals, including the hospitals in the 2nd, 3rd, and 4th quartile, gain patients. In the next step, we divert patients in the second quartile below the threshold and so on. In this scenario 2, more hospitals are able to reach the minimum volume standard at the end of the fourth and final cycle, because some hospitals below the threshold gained enough patients.9 3. RESULTS Estimation results for in-hospital mortality for AAA and HIPs patients are shown in Table III. We present results as odds ratios. The c-statistics above 0.75 for both conditions are in line with other volume-outcome studies or even better (e.g., Birkmeyer et al., 2002; Sund, 2010) and indicate that the models can differentiate well between patients who are alive and patients who died. For both indications, a significant effect of case volume on quality is observable. The case volume quintiles are jointly significant at the 3% level and at the 1% level for AAA and HIP, respectively. Compared with patients with AAA treated in very low-volume hospitals with less than 19 operations per year, all other patients have a lower odds ratio of mortality. Though, those coefficients are significant only for the third and fourth volume quintile. For the fifth quintile, we do not observe a significant difference but this might be because of the low number of hospitals of only 15 in this quintile. The differences between two adjoining quintiles are not significant. For HIP, significant differences to patients treated in hospitals with the lowest case volume exists for patients treated in hospitals with more than 82 cases (as from the third case volume quintile). There are also significant differences between two adjoining quintiles except for the second and third quintile of HIP.

7

We use the user-written Stata command traveltime by Ozimek and Miles (2011). In the USA, the Leapfrog Group (2013) recommends 50 AAA procedures per hospital, whereas in Scotland, at least 20 procedures per year are recommended (Healthcare Improvement Scotland, 2011). No national or international standards are defined for hip fractures. 9 Furthermore, we have performed additional sensitivity analyses by (i) allowing for so-called ‘sole providers’ (e.g., Gale and Coburn, 2003), that is, a hospital stays in the market, even though its volume is below the standard and (ii) by introducing capacity thresholds for remaining hospitals. As these results do not change our results significantly, we refrain from presenting them here. 8



651


Table III. In-hospital mortality estimates (logit) AAA Case volume quintile 2 Case volume quintile 3 Case volume quintile 4 Case volume quintile 5

HIP

Odds ratio

CI

Odds ratio

CI

0.9274 0.6051** 0.5466*** 0.6624

[0.6288,1.3677] [0.3876,0.9446] [0.3502,0.8533] [0.4055,1.0821]

0.9069* 0.8466*** 0.7031*** 0.5976***

[0.8091,1.0165] [0.7526,0.9523] [0.6155,0.8032] [0.5129,0.6964]

1.0678*** 1.8296*** 0.9849 1.5589***

[1.0638,1.0719] [1.7146,1.9523] [0.9141,1.0613] [1.2694,1.9144]

Patient characteristics Age Male Emergency Transfer EVAR Femoral neck fracture CCI: 1–2 3–4 5 Transfer between departments Winter Weekend

1.0765*** 0.8271 1.4307** 1.1889 0.3622***

[1.0556,1.0978] [0.5799,1.1797] [1.0470,1.9552] [0.6010,2.3521] [0.2594,0.5057]

1.9080*** 4.9440*** 7.7027*** 1.0574 1.0008 1.2209

[1.1783,3.0898] [3.0873,7.9172] [4.5025,13.1776] [0.7871,1.4204] [0.7714,1.2986] [0.7741,1.9256]

1.0408 2.1814*** 3.7864*** 6.9337*** 1.8974*** 1.0928*** 1.0115

[0.9838,1.1010] [1.9965,2.3834] [3.4292,4.1808] [6.1776,7.7823] [1.7380,2.0714] [1.0303,1.1591] [0.9518,1.0748]

Hospital characteristics Private non-profit Private for-profit University hospital Teaching hospital Beds: 201–500 > 500 ICU

0.9538 0.8598 1.3939 1.0392 1.3244 1.3259 0.5889***

[0.6758,1.3460] [0.5909,1.2510] [0.8741,2.2228] [0.7327,1.4740] [0.6690,2.6221] [0.6576,2.6732] [0.4308,0.8051]

0.9824 1.0272 1.1287 1.1073** 1.1239** 1.1909** 0.7537***

[0.9043,1.0673] [0.9096,1.1599] [0.8726,1.4601] [1.0109,1.2128] [1.0078,1.2534] [1.0299,1.3772] [0.6918,0.8211]

Urban

0.6632**

[0.4543,0.9681]

0.9775

[0.9028,1.0583]

Observations Number of hospitals Pseudo R-squared C-statistic

7980 386 0.121 0.780

89203 1155 0.113 0.761

Odds ratios with 95% CI in brackets based on cluster-robust standard errors. AAA, abdominal aortic aneurysm; HIP, hip fracture; EVAR, endovascular aneurysm repair; CCI, Charlson comorbidity index. *p < 0.10; **p < 0.05; *** p < 0.01.

Figure 1 shows the estimated average in-hospital mortality probability in each case volume quintile for both conditions with all other variables at their means. We can see a steady decrease of in-hospital mortality with an increasing case volume, except for the fifth case volume quintile for AAA patients. Patients with HIP who are treated in hospitals with less than 58 cases per year have an average probability of death of 5.1% compared with an average mortality of 3.1% for patients who are treated in hospitals with more than 151 cases. Because of the high incidence of HIPs, in total, 364 deaths per year (95% CI, 349–379 deaths) could presumably be avoided if all patients treated in hospitals with less than 58 cases (n D 18468) were treated in an average hospital of the fifth quintile. This corresponds to 6.5% of all deaths for selected HIP patients. For patients with AAA, the case volume effect is lower. However, compared with patients treated in hospitals with less than 15 cases per year, the average probability of death for patients treated in hospitals with more than 68 cases is 1.0 percentage point less. If all patients with AAA treated in hospitals with less than 15 cases per year (n D 1656) would be treated in an average hospital of the fifth quintile, 16 deaths (95% CI, 15–17 deaths) corresponding to 5.7 % of all deaths could presumably be avoided. Copyright © 2014 John Wiley & Sons, Ltd.


652


HIP

5.0 4.5 4.0

4.3

3.5 3.0

3.0

3.1 2.9

2.5

2.0

1.9 1.6

1.0

1.5

1.8

0.0

0.5

0.5 0.0 10−58

1.5

2.5

2.7

2.0

3.5

3.6

Estimated probability of death in %

4.6

4.0

4.5

5.1

1.0

Estimated probability of death in %

5.0

5.5

5.5

6.0

6.0

AAA

59−81

82−109

110−150

Case volume

151−387 3−15

16−25

26−39

40−67

68−209

Case volume

Figure 1. Estimated probability of death with 95 % CI per case volume quintile for abdominal aortic aneurysm (AAA) and hip fracture (HIP) patients

Patient characteristics are basically significant for both AAA and HIP models. It is found that increasing age and increasing number and severity of comorbidities are associated with a significant increase in the odds of mortality for both indications. Not surprisingly, older patients and patients with more comorbidities show a higher risk of death. Furthermore, compared with patients with open aneurysm repair patients with an EVAR have a significant lower odds of mortality. The odds of mortality are significant higher for men compared with women in the HIP treatment. Admissions during winter time or transfers between departments during the hospital stay also increase the odds of mortality for HIP patients. An intensive care unit lowers the odds of mortality for AAA and HIP treatments. HIP patients who are treated in teaching hospitals and hospitals with more beds have a higher odd of mortality compared with patients in non-teaching hospitals or hospitals with less than 200 beds. Tables IV and V show the effects of a hypothetical minimum volume standard on minimum travel times for HIP and AAA patients. Average-minimum travel times for HIP patients (Table IV) are 11 min in status quo with a maximum of 148 min to the nearest hospital. Ninety-five percent of all patients in our sample would reach a hospital within 23 min. In this baseline scenario, all 1155 hospital still provide services. In scenario 1, all 484 hospitals of the first quintile (compare Table II) lose its authorization to treat HIP patients leaving 671 hospitals in the sample. This scenario leads to an increase in average travel time by more than 2 min. The maximum travel time in this scenario would be 221 min with a 95% percentile of 30 min. In comparison with scenario 1, a stepwise introduction (scenario 2) substantially reduces travel times. Average travel times are around 12 min with a maximum time of 190 min. Ninety-five percent of the patients reach the nearest hospital within 26 min. The stepwise introduction also allows more hospitals to reach the minimum volume standard. In comparison with scenario 1, around 100 hospitals more stay in the market, while all 775 remaining hospitals fulfill the minimum volume standard of 58 patients per hospital. ‘Actual’ travel times with all 1155 hospitals are substantially longer. Note that average ‘actual’ travel times are partly driven by outliers as the high standard deviation and maximum in the last column of Table IV shows. Nevertheless, when looking at the outlier robust percentiles, the median driving time with 12 min as well as the 95% percentile with 34 min is longer than in both scenarios with a minimum volume standard. Copyright © 2014 John Wiley & Sons, Ltd.


653


Table IV. Travel times in minutes for HIP Minimum time Mean Standard deviation Minimum Maximum 5% percentile 25% percentile 50% percentile 75% percentile 95% percentile Number of hospitals Number of patients

11.0 6.7 0 148 3 6 10 15 23 1155

Scenario 1

Scenario 2

13.2 9.8 0 221 3 7 11 18 30

12.2 8.2 0 190 3 6 10 17 26

671

775

Actual time 16.6 28.6 0 609 3 7 12 18 34 1155

89203

Minimum time, minimum possible travel times; scenario 1, immediate introduction of minimum volume standard; scenario 2, stepwise introduction of minimum volume standard; actual time, unchanged travel time. HIP, hip fracture.

Table V. Travel times in minutes for AAA Minimum time

Scenario 1

Scenario 2

Actual time

Mean Standard deviation Minimum Maximum 5% percentile 25% percentile 50% percentile 75% percentile 95% percentile

16.9 11.5 0 212 4 9 14 23 38

20.3 14.8 0 212 4 10 17 27 49

18.9 13.0 0 212 4 9 16 26 44

27.7 30.7 0 478 5 11 19 33 72

Number of hospitals Number of patients

386

177

208

386

7980

Minimum time: minimum possible travel times; scenario 1: immediate introduction of minimum volume standard; scenario 2: stepwise introduction of minimum volume standard; Actual time: unchanged travel time. AAA, abdominal aortic aneurysm.

Turning to Table V, it shows that AAA patients have an average-minimum travel time of around 17 min with a maximum of 212 min and 38 min for the 95% percentile. The immediate introduction of a minimum volume standard of 15 patients per hospital (scenario 1) would lead to the exit of 209 hospitals and to an increase of more than 3 min for the average travel time. Ninety-five percent of the AAA patients would reach one of the remaining 177 hospitals within 49 min. A stepwise introduction of the minimum volume standard (scenario 2) would save more than 2 min of average travel time in comparison with scenario 1, while 31 more hospitals would stay in the market. The nearest hospital can be reached in 44 min by 95% of all AAA patients. For AAA patients, ‘actual’ travel times are also substantially longer than both scenarios with minimum volume standards. On average, patients travel nearly 28 min to a hospital, that is, more than 7(9) min longer than in scenario 1 (scenario 2). Again, ‘actual’ mean travel times are driven by outliers. Some patients have exceptional long travel times as the maximum with 478 min shows. Nevertheless, the 50% percentile and especially the 95% percentile show significant longer travel times in comparison with both scenarios. 4. DISCUSSION In this study, we have estimated the effect of case volume on outcome on the basis of administrative data, which includes the full population of all in-patients from Germany from the year 2007. We have shown a Copyright © 2014 John Wiley & Sons, Ltd.


654


significant inverse relationship between volume and mortality for the treatment of patients with AAA and HIP. We provide estimates for the number of lives that could potentially be saved, if patients treated in hospitals of the lowest volume quintile would have been treated in hospitals having the highest number of patients: our estimation results suggest that in this case, between 349 and 379 deaths for HIP patients and 15–17 deaths for AAA patients would be avoidable. Hence, for HIP patients around 6.5% and for AAA patients around 5.7% of all deaths per year are preventable only by choosing a high-volume hospital. We derived hypothetical minimum volume standards for both conditions from our econometric analyses and analyzed changes in travel times for patients using different scenarios. For AAA, the hypothetical minimum volume standard is below international recommendations of between 20 to 50 patients per year. In both scenarios, a significant number of hospitals would lose authorization for treatment, that is, between 380 and 484 hospitals treating HIP patients as well as between 178 and 209 hospitals treating AAA patients would exit the market. However, our findings suggest only a marginal increase in travel burdens for patients, and only if all patients had chosen the nearest hospital for treatment. If we compare travel times under minimum volume standards in scenario 1 or 2 with ‘actual’ travel times, the average travel burden for patients would decrease significantly. Hence, we cannot find any empirical evidence for deteriorating access under minimum volume standards. Furthermore, considering ‘actual’ travel time, the results indicate that some patients prioritize hospitals on other characteristics than travel times. Jung et al. (2011) or Varkevisser et al. (2012) showed, for example, that patients are willing to travel significantly longer to a hospital with a better satisfaction rating. Hence, it is reasonable to assume that patients would be willing to literally ‘go the extra mile’, if they can expect better outcome quality. The present analysis has some limitations. First, our approach of defining minimum volume standard is somewhat arbitrary. We defined minimum volume standards in accordance with volume quintiles of our econometric specification. The derivation of an exact threshold should include further sensitivity analyses (IQWiG, 2005). Second, it remains unclear to what extent treatment of patients with related conditions might suffer due to centralization processes (Birkmeyer et al., 2003b). However, defining related conditions is not easily performed as the literature regarding antitrust and hospital market concentration shows (Gaynor and Vogt, 2000). Furthermore, following the intuition that hospitals with a case volume above the standard provide better services for patients with related conditions without a standard, it may be that outcomes for related conditions improve. Accordingly, outcomes for related conditions might deteriorate, if the patient is treated in a hospital without authorization to treat the primary condition. However, for ruptured aneurysms as the obvious related condition for AAA patients, we do not expect significant changes in treatment. Ruptured aneurysms are emergency cases and require immediate medical attention, so these patients have to seek treatment in the nearest available hospital anyway. Nevertheless, further research should take this potential trade-off into account. Third, we cannot estimate the effects of minimum volume standards on the financial well-being of low-volume hospitals losing authorization to treat patients. However, at least for AAA, the relative small volume numbers for each hospital seem to imply that this issue cannot be a major concern for the financial sustainability of hospitals in Germany. Additionally, Germany has an above OECD average of hospitals and hospital beds (Kumar and Schoenstein, 2013) raising concerns of excess capacities in respect to the number of hospitals (Augurzky et al., 2013). Finally, there are concerns of minimum volume standards setting provider incentives to increase volume above the threshold (Birkmeyer et al., 2003b). However, at least for selected conditions in this study, treatment options seem entirely nondiscretionary, so unnecessary surgery is unlikely. We conclude by emphasizing that minimum volume standards can only refer to elective procedures or procedures with a ‘deferrable priority’ such as HIP patients, that is, adequate access to emergency treatment has to be guaranteed independently. Given the aforementioned limitations and acknowledging the discussed tradeoffs, we are hesitant to recommend any ‘real’ minimum volume standards. More research for both conditions is necessary for (i) deriving potential minimum volume cutoffs and (ii) assessing consequences for related conditions. However, we could show that centralization for AAA and HIP patients improves outcomes with negligible consequences for the average travel burden of patients. Copyright © 2014 John Wiley & Sons, Ltd.


655


APPENDIX A

Table A1. Variable definitions Variable Patient characteristics Mortality EVAR (AAA) Femoral neck fracture (HIP) Age Male Admission Emergency Transfer CCI 0 CCI 1–2 CCI 3–4 CCI 5 Transfer between departments Winter Weekend Hospital characteristics Case volume quintiles AAA Case volume quintile 1 Case volume quintile 2 Case volume quintile 3 Case volume quintile 4 Case volume quintile 5 Case volume quintiles HIP Case volume quintile 1 Case volume quintile 2 Case volume quintile 3 Case volume quintile 4 Case volume quintile 5 Teaching hospital University hospital Public Private non-profit Private for-profit 200 beds 201–500 beds > 500 beds Intensive care unit Urban

Definition 1, if patient died in hospital, 0 otherwise 1, if EVAR, 0 otherwise 1, if femoral neck fracture, 0 otherwise Age in years 1, if male, 0 otherwise 1, if admission by the doctor, 0 otherwise 1, if admission as emergency, 0 otherwise 1, if admission from another hospital, 0 otherwise 1, if Charlson comorbidity index = 0, 0 otherwise 1, if Charlson comorbidity index = 1 or = 2, 0 otherwise 1, if Charlson comorbidity index = 3 or = 4, 0 otherwise 1, if Charlson comorbidity index 5, 0 otherwise 1, if transfer between departments during the hospital stay, 0 otherwise 1, if admission in November, December, January, or February, 0 otherwise 1, if admission on Saturday or Sunday or on public holidays, 0 otherwise Volume Volume Volume Volume Volume

3 and 15 16 and 25 26 and 39 40 and 67 68 and 209

Volume 10 and 58 Volume 59 and 81 Volume 82 and 109 Volume 110 and 150 Volume 151 and 387 1, if teaching hospital, 0 otherwise 1, if university hospital, 0 otherwise 1, if publicly owned hospital, 0 otherwise 1, if private non-profit hospital, 0 otherwise 1, if private for-profit hospital, 0 otherwise 1, if number of beds 200, 0 otherwise 1, if number of beds > 200 and 500, 0 otherwise 1, if number of beds > 500, 0 otherwise 1, if an intensive care unit exists, 0 otherwise 1, if regional structure is urban, 0 otherwise

CCI, Charlson comorbidity index; EVAR, endovascular aneurysm repair; AAA, abdominal aortic aneurysm; HIP, hip fracture. Omitted category.

ACKNOWLEDGEMENTS

We thank Klaus Focke, Uwe Mehlhorn, and Daniel Viehweg from the BKK Federal Association. Furthermore, we thank Boris Augurzky, Christoph M. Schmidt, Harald Tauchmann, and two anonymous referees for helpful remarks as well as Rüdiger Budde for his support with the geocoding of our data. The administrative data of §21 KHEntgG were used as part of a cooperation agreement for the further development of the DRG-system from April 1, 2011 between the RWI and the BKK Federal Association. Roman Mennicken thankfully acknowledges partial financial support by the Dr. Werner-Jackstädt Foundation. Copyright © 2014 John Wiley & Sons, Ltd.


656


REFERENCES

AHRQ – Agency for Healthcare Research and Quality. 2007. AHRQ Quality Indicators - Guide to Inpatient Quality Indicators: Quality of Care in Hospitals - Volume, Mortality, and Utilization, Version 3.1: Rockville. AOK-Bundesverband, Forschungs- und Entwicklungsinstitut für das Sozial- und Gesundheitswesen Sachsen-Anhalt (FEISA) HELIOS Kliniken Wissenschaftliches Institut der AOK (WIdO). 2007. Qualitätssicherung der stationären Versorgung mit Routinedaten (QSR), Abschlussbericht: Bonn. Augurzky B, Krolop S, Hentschker C, Pilny A, Schmidt CM. 2013. Krankenhaus Rating Report 2013 - Krankenhausversorgung zwischen Euro-Krise und Schuldenbremse. Heidelberg: medhochzwei. Ayanian JZ, Weissman JS. 2002. Teaching hospitals , quality of care: a review of the literature. The Milbank Quarterly 346(15): 1128–1137. Barker D, Rosenthal G, Cram P. 2011. Simultaneous relationships between procedure volume and mortality: do they bias studies of mortality at specialty hospitals? Health Economics 80(3): 569–593. Beck A, Rüter A. 1998. Schenkelhalsfrakturen - Diagnostik und therapeutisches Vorgehen. Der Unfallchirurg 101(8): 634–648. Birkmeyer JD, Siewers AE, Finlayson EVA, Stukel TA, Lucas FL, Batista I, Welch HG, Wennberg DE. 2002. Hospital volume and surgical mortality in the United States. The New England Journal of Medicine 346(15): 1128–1137. Birkmeyer JD, Stukel TA, Siewers AE, Goodney PP, Wennberg DE, Lucas FL. 2003a. Surgeon volume and operative mortality in the United States. The New England Journal of Medicine 349(22): 2117–2127. Birkmeyer JD, Siewers AE, Marth NJ, Goodman DC. 2003b. Regionalization of high-risk surgery and implications for patient travel times. The Journal of the American Medical Association 290(20): 2703–2708. Bell CM, Redelmeier DA. 2001. Mortality among patients admitted to hospitals on weekends as compared with weekdays. New England Journal of Medicine 345(9): 663–668. Browne JA, Pietrobon R, Olson SA. 2009. Hip fracture outcomes: does surgeon or hospital volume really matter? The Journal of Trauma 66(3): 809–814. BBSR – Bundesinstitut für Bau, Stadt-und Raumordnung. 2011. Bildung, Gesundheit, Pflege – Auswirkungen des demographischen Wandels auf die soziale Infrastruktur BBSR-Berichte KOMPAKT 11/2011: Bonn. BQS – Bundesgeschäftsstelle Qualiätssicherung. 2008. BQS-Bundesauswertung 2007, Hüftgelenknahe Femurfraktur: Düsseldorf. Charlson ME, Pompei P, Ales KL, MacKenzie CR. 1987. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. Journal of Chronic Diseases 40(5): 373–383. DKG – Deutsche Krankenhausgesellschaft, Spitzenverbände der Krankenkassen (GKV), Verb, der privaten Krankenversicherung (PKV), Institut für das Entgeltsystem im Krankenhaus (InEK gGmbH). 2006. Deutsche Kodierrichtlinien, Allgemeine und spezielle Kodierrichtlinien für die Verschlüsselung von Krankheiten und Prozeduren, Version 2007: Siegburg u.a. Dimick JB, Upchurch GR. 2008. Endovascular technology, hospital volume, and mortality with abdominal aortic aneurysm surgery. Journal of Vascular Surgery 47(6): 1150–1154. Eckstein HH, Bruckner T, Heider P, Wolf O, Hanke M, Niedermeier HP, Noppeney T, Umscheid T, Wenk H. 2007. The relationship between volume and outcome following elective open repair of abdominal aortic aneurysms (AAA) in 131 German hospitals. European Journal of Vascular and Endovascular Surgery 34(3): 260–266. Elixhauser A, Steiner C, Harris DR, Coffey RM. 1998. Comorbidity measures for use with administrative data. Medical Care 36(1): 8–27. Forte ML, Virnig BA, Swiontkowski MF, Bhandari M, Feldman R, Eberly LE, Kane RL. 2010. Ninety-day mortality after intertrochanteric hip fracture: does provider volume matter? The Journal of Bone and Joint Surgery 92(4): 799–806. Gale J, Coburn A. 2003. The Characteristics and Roles of Rural Health Clinics in the United States: A Chartbook. Edmund S. Muskie School of Public Service: Portland. Gandjour A, Lauterbach KW. 2003. The practice-makes-perfect hypothesis in the context of other production concepts in health care. American Journal of Medical Quality 18(4): 171–175. Gaynor M, Vogt WB. 2000. Antitrust and competition in health care markets, In Handbook of health economics, Culyer AJ, Newhouse JP (eds). Elsevier: Amsterdam, Oxford; 1405–1487. Gaynor M, Seider H, Vogt WB. 2005. The volume-outcome effect, scale economies, and learning-by-doing. The American Economic Review 95(2): 243–247. G-BA – Gemeinsamer Bundesausschuss. 2010. Qualitätssicherungs-Richtlinie zum Bauchaortenaneurysma (QBAA-RL) – Anlage 1: Diagnosen (ICD)- und Prozeduren (OPS)-Kodes zum Bauchaortenaneurysma, zuletzt geändert am 11. November 2010: Berlin. G-BA – Gemeinsamer Bundesausschuss. 2013. Regelungen des Gemeinsamen Bundesausschusses gemäß § 137 Abs. 3 Satz 1 Nr. 2 SGB V für nach § 108 SGB V zugelassene Krankenhäuser (Mindestmengenregelung, Mm-R): Berlin. Geraedts M, Cruppé W de, Blum K, Ohmann C. 2008. Umsetzung und Auswirkungen der Mindestmengen – Ergebnisse der Begleitforschung. Deutsches Ärzteblatt 105(51-52): 890–896. Copyright © 2014 John Wiley & Sons, Ltd.



657

Hamilton BH, Hamilton VH. 1997. Estimating surgical volume-outcome relationships applying survival models: accounting for frailty and hospital fixed effects. Health Economics 6(4): 383–395. Hamilton BH, Ho V. 1998. Does practice make perfect? Examining the relationship between hospital surgical volume and outcomes for hip fracture patients in Quebec. Medical Care 36(6): 892–903. Healthcare Improvement Scotland. 2011. Abdominal Aortic Aneurysm (AAA) Screening Programme: Edinburgh. Hessisches Sozialministerium. 2008. Hessischer Krankenhausrahmenplan 2009. Allgemeiner Teil: Planungsgrundlagen und Bedarfsprognose: Wiesbaden. Ho V, Town RJ, Heslin MJ. 2007. Regionalization versus competition in complex cancer surgery. Health Economics, Policy and Law 2(1): 1–71. Ho V, Short MN, Ku-Goto MH. 2012. Can centralization of cancer surgery improve social welfare? Forum for Health Economics and Policy 15(2): 1–25. Hogg RS, Rabout J, Bigham M, Mantaner JS, OShaugnessy M, Schechter MT. 1998. Relation between hospital HIV/AIDS caseload, mortality among persons with HIV/AIDS in Canada. Clinical and Investigative Medicine 21(1): 27–32. Holt PJE, Poloniecki JD, Khalid U, Hinchliffe RJ, Loftus IM, Thompson MM. 2009. Effect of endovascular aneurysm repair on the volume-outcome relationship in aneurysm repair. Circulation: Cardiovascular Quality and Outcomes 2(6): 624–632. Hosmer DW, Lemeshow S. 2000. Applied Logistic Regression (2nd ed.). Wiley: New York. Hughes RG, Garnick DW, Luft HS, McPhee SJ, Hunt SS. 1988. Hospital volume and patient outcomes: the case of hip fracture patients. Medical Care 26(11): 1057–1067. Institute of Medicine. 2000. To Err is Human: Building a Safer Health System: Washington, DC. IQWiG – Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen. 2005. Entwicklung und Anwendung von Modellen zur Berechnung von Schwellenwerten bei Mindestmengen für Knie-Totalendoprothese: Köln. Jung K, Feldman R, Scanlon D. 2011. Where would you go for your next hospitalization? Journal of Health Economics 30(4): 832–841. Kansagra SM, Curtis LH, Schulman KA. 2004. Regionalization of percutaneous transluminal coronary angioplasty and implications for patient travel distance. The Journal of the American Medical Association 292(14): 1717–1723. Keeler EB, Rubenstein LV, Kahn KL, Draper D, Harrison ER, McGinty MJ, Rogers WH, Brook RH. 1992. Hospital characteristics and quality of care. The Journal of the American Medical Association 268(13): 1709–1714. Kumar A, Schoenstein M. 2013. Managing Hospital Volumes - Germany and Experiences from OECD Countries. OECD: Paris. Landon BE, O’Malley AJ, Giles K, Cotterill P, Schermerhorn ML. 2010. Volume-outcome relationships and abdominal aortic aneurysm repair. Circulation 122(13): 1290–1297. Leapfrog Group. 2011. Factsheet: Evidence-based Hospital Referral: Washington, DC. Leapfrog Group. 2013. EBHR Scoring: What does a Hospital’s Overall Score mean? Accessed on 7 November 2013, www. leapfroggroup.org/patients/ebhr_scoring. Luft HS, Bunker JP, Enthoven AC. 1979. Should operations be regionalized? The empirical relation between surgical volume and mortality. The New England Journal of Medicine 301(25): 1364–1369. Luft HS, Hunt SS, Maerki SC. 1987. The volume-outcome relationship: practice-makes-perfect or selective-referral patterns? Health Services Research 22(2): 157–182. McPhee JT, Robinson WP, Eslami MH, Arous EJ, Messina LM, Schanzer A. 2011. Surgeon case volume, not institution case volume, is the primary determinant of in-hospital mortality after elective open abdominal aortic aneurysm repair. Journal of Vascular Surgery 53(3): 591–599.e2. Milcent C. 2005. Hospital ownership, reimbursement systems and mortality rates. Health Economics 14(11): 1151–1168. Ozimek A, Miles D. 2011. Stata utilities for geocoding and generating travel time and travel distance information. Stata Journal 11(1): 106–119. Patti M, Comera C, Glasgow R, Way L. 1998. A hospital’s annual rate of esophagectomy influences the operative mortality rate. Journal of Gastrointestinal Surgery 2(2): 186–192. Quan H, Sundararajan V, Halfon P, Fong A, Burn B, Luthi JC, Saunders LD, Beck CA, Feasby TE, Ghali WA. 2005. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Medical Care 43(11): 1130–1139. Rigberg DA, Zingmond DS, McGory ML, Maggard MA, Agustin M, Lawrence PF, Ko CY. 2006. Age stratified, perioperative, and one-year mortality after abdominal aortic aneurysm repair: a statewide experience. Journal of Vascular Surgery 43(2): 224–229. Romano PS, Chan BK, Schembri ME, Rainwater JA. 2002. Can administrative data be used to compare postoperative complication rates across hospitals? Medical Care 40(10): 856–867. Rommens PM. 2001. Ist die hüftgelenksnahe Fraktur im hohen Lebensalter ein Notfall? Notfall & Rettungsmedizin 4(4): 268–270. Copyright © 2014 John Wiley & Sons, Ltd.


658


Sund R. 2010. Modeling the volume-effectiveness relationship in the case of hip fracture treatment in Finland. BMC Health Services Research 10: 238. Taylor HD, Dennis DA, Crane HS. 1997. Relationship between mortality rates and hospital patient volume for Medicare patients undergoing major orthopaedic surgery of the hip, knee, spine, and femur. The Journal of Arthroplasty 12(3): 235–242. Thiemann DR, Coresh J, Oetgen WJ, Powe NR. 1999. The association between hospital volume and survival after acute myocardial infarction in elderly patients. The New England Journal of Medicine 340(21): 1640–1648. Torsello G, Can A, Schumacher S. 2005. Das Bauchaortenaneurysma. Gefäßchirurgie 10(2): 139–153. Varkevisser M, van der Geest SA, Schut FT. 2012. Do patients choose hospitals with high quality ratings? Empirical evidence from the market for angioplasty in the Netherlands. Journal of Health Economics 31(2): 371–378. Wen SW, Simunovic M, Williams JI, Johnston KW, Naylor CD. 1996. Hospital volume, calendar age, and short term outcomes in patients undergoing repair of abdominal aortic aneurysms: the Ontario experience, 1988-92. Journal of Epidemiology and Community Health 50(2): 207–213. Wenning M, Hupe K, Scheuer I, Senninger N, Smektala R, Windhorst T. 2000. Ist viel gleich gut? Eine Analyse von 116.000 Patienten zum Zusammenhang zwischen Fallzahl und Ergebnisqualität. Der Chirurg 71(6): 717–722.



Minimum cement volume for vertebroplasty.

Minimum effective volume of normal saline for epidural volume extension.

Poverty dynamics in Germany: Evidence on the relationship between persistent poverty and health behavior.

The nebulous relationship between volume and outcome.

A minimum volume covering approach with a set of ellipsoids.

Relationship between Interleukin-6 gene polymorphism and hippocampal volume in antipsychotic-naïve schizophrenia: evidence for differential susceptibility?

His and Hers: Economic Factors and Relationship Quality in Germany.

Catecholamine and volume therapy for cardiac surgery in Germany--results from a postal survey.

The relationship between patient volume and mortality in American trauma centres: a systematic review of the evidence.

Minimum effective volume of combined lidocaine-bupivacaine for analgesic subparaneural popliteal sciatic nerve block.

The relationship between mean platelet volume and varicocele.

The relationship between trauma center volume and in-hospital outcomes.

The relationship between mean platelet volume and pelvic inflammatory disease.

The relationship between thyroid volume and malignant thyroid disease.

The relationship between hospital volume and mortality in severe sepsis.

On the Hospital Volume and Outcome Relationship: Does Specialization Matter More Than Volume?

The relationship between total red cell volume, plasma volume and venous haematocrit.

Development of Cheaper Embryo Vitrification Device Using the Minimum Volume Method.

The vascular volume-pressure relationship in the dogfish, Squalus acanthias.

Dynamics of an axisymmetric liquid bridge close to the minimum-volume stability limit.

The Relationship between Medicare Payment and Service Volume for Retina Procedures from 2005 through 2009.

Social Bonds and Exercise: Evidence for a Reciprocal Relationship.

Potassium intake and stroke risk: a review of the evidence and practical considerations for achieving a minimum target.

The effects of walking speed on minimum toe clearance and on the temporal relationship between minimum clearance and peak swing-foot velocity in unilateral trans-tibial amputees.