WHAT CAN WE LEARN FROM A CROSS-COUNTRY COMPARISON OF THE COSTS OF CHILD DELIVERY?

Transcription

1 HEALTH ECONOMICS Health Econ. 17: S47 S57 (2008) Published online in Wiley InterScience ( WHAT CAN WE LEARN FROM A CROSS-COUNTRY COMPARISON OF THE COSTS OF CHILD DELIVERY? MARTINE M. BELLANGER a and ZEYNEP OR b, * a E cole Nationale de la Sante Publique, Rennes, France b Institut de Recherche et Documentation en Economie de la Sante, Paris, France SUMMARY This study provides a comparative analysis of the costs of normal delivery in hospital in nine European countries using the data collected as part of the HealthBASKET project. The results show that both the level of input (medical labour) prices and the skill mix used for delivery are major determinants of total delivery costs. At the hospital level, there seems to be room for greater efficiency through specialisation and task shifting from doctors to midwives and nurses. More generally, the results of our study suggest that the costs of delivery in hospital are not independent of supplementary home care provided outside of hospitals. The cost information and analysis in this study may also be useful for developing healthcare-specific purchasing power parities (PPPs) that allow for healthcare expenditures to be compared across countries. Copyright # 2008 John Wiley & Sons, Ltd. KEY WORDS: delivery; healthcare costs; costs and cost analysis; regression analysis INTRODUCTION Childbirth is one of the most important events in people s lives, and it also represents a significant proportion of healthcare expenditures. In the industrialised nations, hospitals are increasingly regarded as the safest and most appropriate place to give birth, whereas in some parts of the world, home birth is still the norm. In most European countries, child delivery accounts for approximately 5% of total hospital activity (OECD, 2006). New technologies have been developed that allow for better monitoring of both mother and foetus during pregnancy and delivery, thus contributing to a reduction in both maternal and neonatal mortality. While international comparisons are often made of the outcomes of medical interventions for child delivery, comparative data on the costs of delivery in hospital are scarce. The present study provides an analysis of the costs of child delivery in hospital across the nine European countries involved in the HealthBASKET project (i.e. Denmark, England, France, Germany, Hungary, Italy, The Netherlands, Poland, and Spain) using a new methodology for comparing hospitalisation costs within and across countries. To improve the production process and pricing of health services, it is important to understand the underlying production process of different treatments and the variety of factors that can drive costs. For the purpose of comparing hospital-care services and their costs across countries, birth by normal delivery provides a particularly good benchmark case, because both the care provided and the outcomes are easy to standardise. The aim of our study is to compare the costs of an episode of care (i.e. a normal delivery) in the hospital setting. This differs from a cost of illness approach, the aim of which would be to estimate the total costs of normal delivery while taking into account care provided before and after delivery in *Correspondence to: Institut de Recherche et Documentation en Economie de la Santé, 10 rue Vauvenargues Paris, France. or@irdes.fr Copyright # 2008 John Wiley & Sons, Ltd.

2 S48 M. M. BELLANGER AND Z. OR hospital. More specifically, this paper has two objectives: (1) to compare the costs and prices (reimbursement rates) of normal delivery across countries and (2) to explore the main factors behind variations in costs across providers within and among the countries studied. In the following section, we will provide some medical background information on normal delivery in hospital, as well as a brief review of the literature. In the third section, we will describe the methods used to collect data and will present descriptive statistics. In the fourth section, we will explore the determinants of hospital delivery costs within and across countries using a series of regression analyses. In the fifth and final sections we will draw a number of conclusions based on our findings. NORMAL CHILD DELIVERY AT HOSPITAL: MEDICAL BACKGROUND AND LITERATURE In all of the countries that participated in our study, the majority of births take place in hospital. However, the exact proportion varies widely between countries, ranging from approximately 60% in the Netherlands to virtually 100% in France and Denmark (Anthony et al., 2005; ONS, 2007; Ankjaer- Jensen and Johansen, 2006). In this study we follow the vignette approach taken by the HealthBASKET project, focusing on normal delivery (i.e. a non-assisted vaginal birth without any complications; see the following section) as an episode of care. This approach helps to improve the comparability of treatment patterns across providers. Typically, an expectant mother presents at the obstetrics department with contractions. On admission, she is almost always examined by a midwife rarely by an obstetrician to determine the status of the uterus, and the position and presentation of the baby. Some laboratory analyses may also be performed, such as complete blood count and haemoglobin tests. The first phase of delivery, called labour, usually takes between 5 and 15 h for a first baby. During this period, a midwife or another skilled nurse provides assistance, and various kinds of medical interventions may take place. For instance, foetal heartbeat and uterine contractions may be monitored by means of cardiotocogram. Medication may be administered to induce labour or reduce the risk of postpartum haemorrhage (de Abajo et al.,2004).paincontrolmayalso be necessary during labour. Antispasmodics are frequently administered, and epidural anaesthesia may be performed if requested by the mother, although different countries show large variations in the frequency of the latter procedure. According to Blondel et al. (2005), 63% of women who gave birth in France in 2003 received an epidural compared with 33% in England in During the last stage of delivery (expulsion), an episiotomy may be performed, although this procedure also shows substantial variations depending on country and healthcare provider. From a strictly medical perspective, an overnight or longer stay in hospital following birth is not necessary for uncomplicated deliveries (Escobar et al., 2001). However, in cases where an epidural or episiotomy has been performed, an observation period of 24 h is usually recommended (Brumfield et al., 1996). Nevertheless, the length of hospital stay can also differ depending on the availability of organised home-care services for the mother and child and on national guidelines for maternal care. The literature on the hospitalisation costs of normal delivery is not extensive. Most studies have examined the cost-effectiveness of different birthing methods for low-risk mothers, comparing, for example, home births managed by nurse midwives to hospital births (Anderson and Anderson, 1999; Palencia et al., 2006) or analysing the economic implications of delivery methods, e.g. caesarean delivery in labour compared with assisted vaginal delivery (Henderson et al., 2001; Allen et al., 2006). Among the very few studies that have examined the determinants of costs associated with hospital births, length of stay was identified as a major source of cost variation (Allen et al., 2005; Schmitt et al., 2006). It has also been shown that the delivery of a first baby costs more than the delivery of a second (Allen et al., 2006; Le Vaillant and Pouvourville, 2006). In our extensive review of the literature, we were unable to find any studies that compared the costs of normal delivery across countries.

3 CROSS-COUNTRY COMPARISON OF THE COSTS OF CHILD DELIVERY S49 DATA FOR NORMAL DELIVERY Sample selection and data collection As part of the HealthBASKET project, a vignette approach was developed for estimating and comparing the costs of different health services at the micro level in a range of countries. The vignettes ensure that patient characteristics such as age and medical condition are standardised for cost/price comparisons, thus avoiding unintended variations in resource use. The vignette applied in our study for normal delivery in hospitals is given in Box I. Box I. Vignette for normal delivery The vignette starts at the hospital door with the admission of a healthy pregnant year-old woman with labour pains after 39 weeks of a first pregnancy without complications. Upon examination, the foetal presentation is normal (i.e. cephalic/vertex; one foetus/single baby, head down). A normal vaginal delivery is performed without complications i.e. the expectant mother spontaneously delivers a mature child of normal weight and there is no subsequent transfer to the paediatric department or newborn intensive-care unit. The vignette ends with the discharge of mother and child, both in good health. Data were collected and analysed by researchers in each country. Costs were calculated from the provider perspective. In each country, at least five representative healthcare providers relevant to the case vignette were identified. Providers whose cost structure was likely to differ considerably from other providers (e.g. university hospitals) were excluded from the sample. Altogether, data for delivery were collected from a sample of 47 hospitals in nine countries, and the number of providers per country ranged from 4 to 11. More details on the selection criteria and sample characteristics in each country are provided in Mason et al. (2007). Although the participating researcher agreed upon a common framework for the type of data to be collected for the analysis, different research teams adopted different approaches to carrying out data collection. All countries obtained data from hospital accounting departments. In addition Italy and Spain conducted face-to-face interviews with health professionals and financial managers; standardised questionnaires were sent out in Germany and Poland; and supplemental data from national cost databases were obtain in Denmark, France, and England. Members of medical and nursing staff were asked to provide information on the resources used at the patient level based on the last 10 patients who matched the standardised description in the case-vignette. However, in some countries, the number of patients covered was considerably higher due to the use of regional or national administrative databases (Mason et al., 2007). In all countries, information was collected on diagnostic procedures before delivery; total costs of drugs administered; total personnel costs in the delivery room and in the ward; and total overheads. Data on unit costs and quantities consumed were requested separately and were provided in most cases. In addition, data were collected on various hospital characteristics such as the total number of beds and physicians. Descriptive analysis Hospital characteristics. Table I shows the main characteristics of the hospitals included in the analysis. The total number of beds per hospital, on average in each country, ranged from approximately 400 in Spain to more than 700 in Denmark and England. The average number of full-time equivalent (FTE) physicians per hospital bed was 0.4, with little notable variation between countries, whereas FTE nurses per bed ranged from more than 4 nurses per bed in the Netherlands to 0.6 in Germany, with an average across the entire sample of 1.7 nurses per bed. The nurse-to-physician ratios varied greatly, ranging from approximately two in Italy and Poland to more than nine in the Netherlands. Regarding the size of

4 S50 M. M. BELLANGER AND Z. OR Table I. Mean characteristics of hospitals and length of stay by country Hospital level Department level Country Hospitals per country Number of beds Capacity utilisation Number of physicians Number of nurses Nurse/physician ratio Number of physicians per bed Number of nurses per bed Number of beds Number of physicians Length of stay (in days) Denmark (200.7) 0.97 (0.05) 411 (88) (405.6) 3.1 (0.3) 0.52 (0.03) 1.62 (0.09) 52 (25.1) 39 (14.9) 2.08 (0.2) England (214.3) 0.80 (0.04) 247 (102.9) (252.4) 4.7 (0.9) 0.35 (0.05) 1.66 (0.3) NA NA 1.74 (0.1) France (196.6) 0.85 (0.05) 137 (54.5) 597 (233.2) 4.35 (0.7) 0.32 (0.05) 1.41 (0.2) 34 (4.8) 5 (0.8) 4.9 (0.4) Germany (210) 0.70 (0.1) 99 (60.7) 241 (135.7) 2.42 (0.6) 0.22 (0.05) 0.54 (0.1) 46 (43.9) 9 (7.1) 3.51 (0.9) Hungary (183.3) 0.79 (0.05) 75 (50.2) 419 (112.2) 5.59 (2.4) 0.14 (0.04) 0.79 (0.03) 27 (2.1) 10 (1.3) 4.38 (1.1) Italy (274.4) 0.89 (0.1) 473 (200.4) (469.7) 2.15 (0.2) 0.68 (0.04) 1.47 (0.14) 34 (9.4) 17 (4.5) 2.6 (0.6) Netherlands (287.8) 0.56 (0.07) 265 (267) (2370) 9.24 (2) 0.47 (0.23) 4.39 (1.9) 18 (6.1) 10 (4.6) 0.86 (0.2) Poland (245) 0.73 (0.05) 127 (70.8) 286 (117) 2.26 (0.9) 0.28 (0.02) 0.63 (0.23) 51 (22.8) 11 (7.9) 4.28 (1.2) Spain (275.6) 0.89 (0.03) 279 (203.8) NA NA 0.69 (0.09) NA NA NA 2.9 (0.6) Total (252.8) 0.78 (0.13) 230 (190) 966 (1221) 4.30 (2.9) 0.43 (0.2) 1.73 (1.3) 36 (12.8) 14 (11.4) 2.95 (1.4) Note: Standard deviation in parentheses.

5 CROSS-COUNTRY COMPARISON OF THE COSTS OF CHILD DELIVERY S51 Table II. Total cost, cost components, and reimbursement of normal delivery Personnel cost Overhead cost Procedure cost Drug cost in h in % in h in % in h in % in h in % Average total cost Reimbursement a Denmark } England France Germany Hungary Italy Netherlands Poland Spain NA Average a DRG prices for a normal delivery without complications. the obstetrics wards, the Netherlands had the smallest (18 beds) and Denmark and Poland the largest (more than 50). Because data on the number of nurses in the participating obstetric wards were only available for a few providers, these were not included in our analysis. Average length of stay. The average length of stay varied from 0.84 days in the Netherlands to 4.9 days in France, with large within-country variations in Poland, Hungary, and Germany (Table I). Both the organisation of the healthcare system and cultural factors are likely to have an important influence on the medical approach taken to normal delivery and related care. Roughly, we can distinguish between countries according to two main models. The first can be described as the home-based model, in which the hospital stay is very short and there is more-or-less regular organised home-nursing aid for the mother and child after delivery. The extreme example of this model is the Netherlands, where the hospital stay is restricted to the time required for delivery (length of stay less than a day) and the entire postpartum delivery period is spent at home with substantial home-care assistance. The second model is that of the hospital-led birth, in which the average length of hospital stay is longer (i.e. between 3 and 5 days) and most postpartum care is provided at the hospital itself. One of the more extreme examples of this model is France, where postpartum care has been standardised with medical protocols (national perinatal plan to reduce neonatal mortality rates) requiring approximately five days of inpatient care for mother and child. 1 The other countries in our sample are located somewhere between these two extremes (Table I). Costs. The cost data were collected in national currencies and converted into euros using 2005 exchange rates. Table II presents the overall cost of a normal hospital delivery, as well as major cost components (i.e. as country means), for the nine countries. It also provides the average reimbursement rates (DRG prices) for normal delivery in each country, except for Spain where global budgets are used for hospital funding. As shown in Table II, the costs of hospital delivery ranged from h342 in Hungary to h2365 in Germany, with an average of roughly h1260 for all nine countries. When interpreting the data presented in Table II, it is important to bear in mind that these are the average total costs of normal delivery for a standardised patient profile, as described in the vignette above, and should not be taken as the average cost of delivery in these countries. The reimbursement rate presented, however, is the average (DRG) price set in these countries for vaginal delivery without complication and assumes an average cost structure (i.e. there will be some under payment for some patients and overpayment for some others). 1 It is worth noting that France, together with Germany, has one of the lowest neonatal mortality rates within the OECD countries (OECD, 2006)

6 S52 M. M. BELLANGER AND Z. OR Table III. Physician and nursing time, unit cost by category of personnel Total care time Physician time Nursing time Average unit labour cost a Unit physician cost Unit nursing cost Country (h) (h) (h) (h=h) (h=h) (h=h) Denmark England France Germany Hungary Italy Netherlands Poland Spain Average a Weighted average of personnel cost (midwife, nurse, and physician) by the time spent per normal delivery. With regard to our vignette, it would appear that the reimbursement rates in most countries correspond quite closely to the average costs of normal delivery, with the exception of England, where the reimbursement rate specified for delivery is 40% lower than the observed average cost. There is also substantial variation across countries in the proportion of the two main cost components: personnel costs vs overhead costs. Personnel costs include physician costs (mainly for obstetricians, but also for anaesthetists and, if applicable, paediatricians), midwife costs, and other nursing costs. For instance, the proportion of total costs attributable to personnel costs was more than 70% in Germany, but only approximately 25% in Italy and Denmark. In terms of overhead costs, it should be noted that measuring overheads in a standardised manner proved to be very difficult in this project because of the differences in accounting rules used to allocate total costs between direct and indirect costs across countries (Mason et al., 2007). Medical tests (procedures) and drugs (including anaesthesia) accounted for a small part of the total costs, ranging from less than 1% in Denmark and England to almost to 10% in Italy and Poland (Table II). Resource use and labour prices for delivery. The main source of variation in delivery costs was total personnel costs, which were directly influenced by the national delivery practices. While some countries primarily employed midwives to provide support to the mother before, during, and after delivery (Denmark, France, and England), in other countries (Germany and Spain) an obstetrician was almost always present during delivery (representing 17 and 6% of the total costs, respectively). In Italy, Poland, and the Netherlands, an obstetrician was in attendance during delivery in some, but not all, hospitals. In turn, these costs can be divided into a price effect and a quantity effect. Thus, the variation in total personnel costs across countries/providers can be explained by differences in the hourly compensation of medical personnel (price) and in their labour time per delivery (quantity). The amount of time spent by medical and nursing staff with mother and child varied significantly across countries (Table III). Clearly, the total time is higher if the stay is longer, but the relationship is not linear as most of the care given is concentrated in the first day. For example, it ranged from approximately 25 h in France for an average length of stay (ALS) of 4.9 days, 10 h in the Netherlands for an ALS of 0.86 days, and 13 h in England for an ALS of 1.7 days. The labour cost of nursing is a weighted average of nurse and midwife costs, and the physician cost is an average of obstetrician, anaesthetist, and/or paediatrician (as applicable). The overall average labour cost for delivery was calculated by weighting the labour cost of each category of personnel by their time inputs (in hours). This yielded an average labour cost of approximately h30 for the sample, ranging from h5 in Poland to h51 in England (Table III). Thus, the average personnel costs per delivery in a country might be high depending on whether labour compensation rates are high or the quantity of resources used (staff time) is high. For example, in England average personnel costs per delivery were

7 CROSS-COUNTRY COMPARISON OF THE COSTS OF CHILD DELIVERY S53 approximately h110 greater than the sample average. However, if personnel time per delivery in England had been equal to the sample mean, the costs would have been more than h400 greater than the sample average. Conversely, in Hungary, where the overall personnel costs were almost h450 lower than the sample average, if the wages for personnel had been set equal to the sample mean, overall costs would have been close to the average. This means that all of the differences in cost stemmed from prices and not from differences in resource use. ANALYSIS OF COST VARIATIONS AND MAIN COST DRIVERS This section explores the relative contribution of major factors that might contribute to variations in total costs, as well as the main cost drivers within and across countries (i.e. labour compensation rates, average length of stay, and staff time per delivery). Model specification We used a multilevel modelling approach that accommodated the hierarchical nature of our data; in other words, data on hospitals (level 1) were nested within country (level 2) groups (Rice and Jones, 1997; Grieve et al., 2004). In its simplest form, the basic model for exploring variations in costs/resource use across and within countries was specified as follows: Y hi ¼ b 0i þ g p x pih þ e hi ð1þ where Y hi corresponds to the dependent variable (in our case, this was unit delivery cost, time input, and labour compensation rate), with the subscript h for hospital ðh ¼ 1;...; 47Þ and i for country ði ¼ 1;...; 9Þ; x pih represents the set of p explanatory variables at the hospital level and g p their associated slope coefficients; b 0i is the intercept varying across countries; and e hi is the residual error term for hospital h in country i: Note that in the above equation (and in the present analysis) the slope coefficients of all the explanatory variables are treated as fixed. Across countries, the intercepts b 0i have a normal distribution with a given mean ðb 0 Þ and variance ðs 0i Þ: b 0i ¼ b 0 þ m 0i ð2þ Moreover, covðm 0i e ih Þ¼0: At level 2, we can also define several more complex variance structures to model more precisely some relationships of interest, by introducing explanatory variables at the country level, as follows: b 0i ¼ b 0 þ az i þ v 0i ð3þ with z i representing a regressor that varies across countries, but fixed for provider units within a country (such as GDP). The terms b 0 and a represent the fixed elements, and v 0i is a random error component assumed to be normally distributed with 0 mean and a variance ðo 0i Þ: The variance components are now conditional (holding z constant) since they represent the variability b 0i after controlling for the countrylevel variables. We can distinguish between two groups of explanatory variables: at the hospital level and the country level. At the hospital level, we have information on the average length of stay for normal delivery (in days), as well as direct and indirect time spent with patients during delivery by physicians and nurses (in minutes). In addition, we introduced three variables to control for hospitals general characteristics: total number of beds in the hospital and the obstetrics ward, the ratio of FTE number of nurses to physicians at the hospital, and the percentage of occupied beds. The first two variables control for the

8 S54 M. M. BELLANGER AND Z. OR Table IV. Estimates of the determinants of costs and major cost drivers of delivery Unit delivery cost b Average labour compensation Care time LOS Dependent variable ðhþ 1 (h; logged) 2 ðh=minþ 3 (min) 4 (days) 5 Intercept nn (258.2) (4.0) (0.19) (673.25) (0.67) GDP nn } } (0.4) ( ) } } Home care } } :4369 nn ( ) (0.93) Hospital beds :00014 n } } (0.102) ( ) Obstetric beds } 0:0022 nn nn (0.0009) (4.15) (0.005) Care time } } } } Nurse/physician } 0:01981 n (0.0147) (73.70) (0.08) ALOS 0:58 nn } } } (0.17) Physician compensation } } } (0.14) Nurse compensation 0.46 nn } } } (0.21) Average labour compensation } } 1088:54 nn (729.31) (0.76) Random effects Variance component Country a Hospital N Deviance Standard deviations are in parenthesis. nn Significant at 1%, n significant at 5% level. a p-value 50:0001 for all variance terms. b All the variables used in column 2 are logged, so that the coefficients can be read as elasticities. Deviance is twice the negative log-likelihood value associated with the maximum likelihood estimates. It indicates the fit of the model to the data (the larger the deviance, the poorer the fit). impact of variations in hospital size and the available labour resources, and the last variable is included because it may influence the length of stay and the time spent by medical staff with patients. At the country level, we used GDP per capita as a proxy variable to control for the impact of overall country differences in the level of living. Moreover, as discussed above, the country-specific organisation of healthcare including, in particular, the availability of organised home care may have an impact on delivery practices and on the length of stay in a hospital. Therefore, we created a dummy variable that takes values between 1 (for the Netherlands) and 0 (France and Germany) to capture the availability of home-nursing care in each country. 2 Regression results Table IV presents the most pertinent results from our exploratory analysis. In the first two columns, the dependent variable is the unit delivery cost measured in euros, using exchange rates from Column 1 presents the results from the unconditional model with no explanatory variables. The 2 The values for the other countries are 0 for Spain and Hungary, 0.25 for Italy and Poland, 0.50 for Denmark and England. These proxy variables were constructed by the authors based on qualitative information from country experts.

9 CROSS-COUNTRY COMPARISON OF THE COSTS OF CHILD DELIVERY S55 preliminary information from the unconditional model is useful for separating total variations in hospital costs into variations between and within countries. We can see that most variation (more than 70% of estimated variance) in costs occurred at the country level (i.e. between countries). The standard deviation from the mean (h1276) across countries was approximately h750, whereas within countries it is approximately h420. The equation in column 2 is a unit cost function for delivery. Unit delivery costs are estimated as a function of factor prices, i.e. nurse and physician labour compensation rates (euros/minute), average length of stay (quantity), and the total number of hospital beds (scale effect). At the country level, GDP per capita was introduced to control for non-labour prices (of equipment, infra-structure, etc.). Following the tradition of using long-term ad hoc cost functions to explain variations in unit costs (Adam et al., 2003) and given the distribution of our cost data, all the variables are expressed in logs. The log transformation has the advantage that, in column 2, coefficients can be interpreted as elasticities. The results suggest that average length of stay and nurse compensation levels were the major determinants of unit delivery costs in our sample. A 10% increase in the length of stay implies a roughly 6% increase in costs. Controlling for nurse compensation levels, the level of physician compensation did not have a significant impact on costs, which underscores the nurse-intensive nature of delivery. Hospital size (number of beds) and bed occupation rates do not seem to have had a significant impact on costs. The results with the occupation rates are not presented in the table because they were not significant in any of the models. The determinants of the main factors driving unit delivery costs are explored in columns 3 5. Note that the equations in these columns are linear in form and are estimated based on a sample of 37 hospitals because of missing data for some providers. In column 3, we look at the factors that might have had an impact on labour compensation costs (euros per minute). As expected, differences across countries in labour costs were closely linked to differences in GDP per capita. At the hospital level, the ratio of nurses to physicians appears to have had a negative impact on total labour costs per delivery. This is not surprising because average labour compensation for nurses is always lower than for doctors. A high nurse/doctor ratio may suggest a higher utilisation of nurses/midwives for tasks otherwise carried out by physicians/obstetricians. More surprisingly, the coefficient of obstetric beds suggests that in hospitals where the number of beds at the ward level is high, labour costs are lower. This might imply a specialisation effect in other words, where there is a reasonably sized obstetric ward, labour resources are used more efficiently. On the other hand, the overall size of the hospital (hospital-beds) has the opposite effect (less efficient use of resources). In column 4, we estimate the determinants of personnel time per delivery (in minutes). In addition to explanatory variables for the number of beds in obstetric wards, the ratio of nurses to doctors, and labour compensation rates, we introduced our dummy variable for the availability of home-care. The only significant factor was the level of labour compensation; the higher the labour costs, the shorter the time spent with patients. In column 5, we estimate the determinants of the average length of stay (in days) using the same explanatory variables as in column 4. In countries where there was extensive post-delivery home care, the length of stay was shorter. At the hospital level, the size of the obstetric ward (beds) appears to have been the only significant determinant of length of stay; in hospitals where the number of beds in the obstetric ward was higher, the length of stay was longer. CONCLUSION Cross-country information on the costs of individual healthcare services is rare. This study provides a comparative analysis of the cost of normal delivery in hospital in nine European countries using the

10 S56 M. M. BELLANGER AND Z. OR data collected as part of the HealthBASKET project. The data collected in this project allow the impact of the prices and volumes on costs of child delivery to be delineated both within and across countries. Nonetheless, we advise caution when interpreting the results presented in this paper, as our study has a number of limitations most of which are related to problems with data collection. For example, in practice, participating researchers found it difficult to collect data on unit prices and direct time spent with mother and child. Therefore, for some countries, certain cost components include estimates. More generally, the vignette approach has its own structural and methodological limitations, which have been discussed elsewhere (Busse et al., 2007). Also, our approach sought to collect only standardised information from a small number of providers. From an econometric point of view, it would have been preferable to have a larger number of providers in each country. Despite these limitations, our analysis demonstrates that there are significant variations across hospitals and countries in the quantity and type of medical resources used (labour) and in the care provided for a standard intervention like delivery. In some hospitals, an obstetrician was always involved in delivery, while, in others, midwives and nurses provided support for the mother before, during, and after delivery. The length of hospital stay ranged from less than 1 day in the Netherlands to about 5 days in France and Germany. The average cost of a normal first delivery at hospital for mothers aged in the nine countries explored in our study was approximately h1260. However, there were significant differences in average costs within and between countries, ranging from roughly h350 in Hungary to more than h2000 in Germany and France. The results of our exploratory analysis show that both the level of input (medical labour) prices and the skill mix used for delivery are major determinants of total delivery cost. The data produced by the HealthBASKET project allowed us to measure and compare the impact of direct care volume (i.e. the time spent by medical personnel) on costs. It appears that, for countries and providers that employ more nurses or midwives in lieu of obstetricians, the cost of delivery is lower. Also, there is a strong negative relationship between time spent with patients and the overall unit labour cost for delivery. Country differences in medical labour costs are also clearly related to national differences in the overall standard of living. The cost information collected for this study may also be useful in developing healthcarespecific purchasing power parities, which would allow for healthcare expenditures to be compared across countries while taking into account differences in price levels of specific health services between countries (Schreyo gg et al., 2007). Our results also have implications for the organisation of child delivery care at the provider and country levels. Indeed, at the hospital level there seems to be room for greater efficiency through specialisation and task shifting from doctors to midwives and nurses. More generally, the results of our study suggest that the cost of delivery in hospital is not independent of the supplementary home care provided outside of hospitals. Thus, in future work, it would be interesting to explore the quality aspects, as well as the cost implications, of different settings for child delivery. ACKNOWLEDGEMENTS The results presented in this article are based on the project Health Benefits and Service Costs in Europe HealthBASKET, which was funded by the European Commission within the Sixth Framework Research Programme (grant no. SP21-CT ). No conflicts of interest declared. CONFLICT OF INTEREST

11 CROSS-COUNTRY COMPARISON OF THE COSTS OF CHILD DELIVERY S57 REFERENCES de Abajo FJ, Meseguer CM, Anton ilo G et al Labor induction with dinoprostone or oxytocine and postpartum disseminated intravascular coagulation. A hospital-based case control study. American Journal of Obstetrics and Gynaecology 191: Adam T, Evans DB, Murray CJL Econometric estimation of country-specific hospital costs. Cost Effectiveness and Resource Allocation 1: 3 (Available from: 19 October 2007). Allen VM, O Connell CM, Baskett TF Cumulative economic implications of initial method of delivery. Obstetrics and Gynaecology 108: Allen VM, O Connell CM, Farrell SA, et al Economic implications of method of delivery. American Journal of Obstetrics and Gynaecology 193: Anderson RE, Anderson DA The cost-effectiveness of home-birth. Journal of Nurse-Midwifery 44: Ankjaer-Jensen A, Johansen JT The cost of ten case-vignettes in the Danish health sector. Results from a survey based on figures from Report Part II for the HealthBASKET project (SP21-CT ). Anthony A, Amelink-Verburg MP, Jacobusse GW et al De thuisbevalling in Nederlands PRN/ TNO rapport. Blondel B, Supernant K, Mazaubrun du C et al Enqueˆte pe rinatale 2003 en France, Situation en 2003 et e volution depuis Rapport pour le Ministe`re de la sante, Paris. Brumfield CG, Nelson KG, Stotser D et al hour mother infant discharge with a follow-up home health visit: results in a selected medicaid population. Obstetrics and Gynaecology 88: Busse R, Schreyo gg J, Smith P Editorial: variability in healthcare treatment costs amongst nine EU countries results from the HealthBASKET project. Health Economics 17(S1): S1 S8. Escobar GJ, Braveman PA, Ackerson L et al A randomized comparison of home visits and hospital-based group follow-up visits after early postpartum discharge. Pediatrics 108: Grieve R, Nixon R, Thompson SG et al Using multilevel models for assessing the variability of multinational resource use and cost data. Health Economics 14: Henderson J, McCandlish R, Kumiega L et al Systematic review of economic aspects of alternative modes of delivery. British Journal of Obstetrics and Gynaecology 108: Le Vaillant M, Pouvourville de G Analyse du couˆt hospitalier public de l accouchement par voie normale. French hospital cost study. Rapport DREES. Linna M Measuring hospital efficiency with panel data models. Health Economics 7: Mason A, Epstein P, Smith PC et al International comparison of costs: an exploration of within and between country variation for ten health care services in nine EU members states. Final report for phase III of HealthBASKET project. (SP21-CT ). National Statistics On Line (ONS). The U.K., Organisation for Economic Co-operation and Development (OECD) OECD Health Data. OECD: Paris. Palencia R, Gafni A, Hannah ME et al The costs of planned caesarean versus planned vaginal birth in the Term Breech. Canadian Medical Association Journal 174: Rice N, Jones A Multilevel models and health economics. Health Economics 6: Schmitt SK, Sneed L, Phibbs CS Costs of newborn care in California: a population-based study. Pediatrics 117: Schreyo gg J, Tiemann O, Stargardt T, Busse R Cross-country comparison of costs: the use of episode-specific transitive purchasing power parities with standardised cost categories. Health Economics 17(S1): S95 S103.