London, Brunei Gallery, October 3 5, The Dutch experience in measuring health output and labour productivity

Transcription

1 Sessin Number : 2 Sessin Title : Health recent experience in measuring utput grwth Sessin Chair : T. ATKINSON Paper prepared fr the jint OECD/ONS/Gvernment f Nrway wrkshp Measurement f nn-market utput in educatin and health Lndn, Brunei Gallery, Octber 3 5, 2006 The Dutch experience in measuring health utput and labur prductivity Antni CHESSA, Statistics Netherlands Fske KLEIMA, Statistics Netherlands Fr additinal infrmatin, please cntact : Authr name(s) : Antni CHESSA Authr adress(es) : Statistics Netherlands (CBS), P.O. Bx 4000, 2270 JM Vrburg, the Netherlands Authr (s) : acsa@cbs.nl Authr telephne(s) : +31 (0) Authr fax(es) : +31 (0) This paper is psted n the fllwing website :

2 The Dutch experience in measuring health utput and labur prductivity Antni G. Chessa 1 and Fske J. Kleima Statistics Netherlands (CBS), Vrburg, the Netherlands Abstract Labur prductivity f health services has received a lt f attentin in the Netherlands in recent years. Labur prductivity requires the measurement f utput and labur input. This paper describes the utput vlume index develped at Statistics Netherlands fr hspital health services and sme f the prblems that are encuntered in the measurement f utput. Next, pssibilities will be discussed fr including ther health sectrs, such as nursing hmes and hmes fr the elderly. The greatest challenge lies in finding a methd that takes int accunt the yearly changes in the cmpsitin f prducts, as sme sectrs shw a strng increase f new extramural services since On the ther hand, the main prducts f each sectr can be tracked frm year t year and their vlumes and prductin values are knwn. This culd serve as a basis fr an utput methd that includes hspitals and ther sectrs. This paper illustrates the utput methd fr hspitals with an applicatin t labur prductivity f general hspitals in the Netherlands. 1. Intrductin The measurement f the prductin vlume f health services is becming increasingly imprtant. The demand fr health services is expected t increase due t factrs like ppulatin ageing and technlgical prgress. A grwth f health services can be achieved in different ways, fr instance, by investing in input factrs, such as capital and labur, and/r by enhancing prductivity with regard t ne r mre inputs, such as labur prductivity. At present, there is a lt f attentin in Dutch plitics fr the latter apprach. Fr the verall labur prductivity f the Netherlands, the health sectr plays an imprtant rle. 2 Apprpriate methds fr quantifying the yearly develpment f utput and labur input are therefre needed. Finding an adequate measure fr utput vlume is nt straightfrward. This task invlves the partitining f ttal utput int a set f individual prducts (i.e., setting criteria in rder t define these prducts), quantifying their vlume and finding weights in rder t aggregate these vlumes int a value representing ttal utput vlume. In rder t distinguish health prducts we need t establish when tw prducts are the same. In this respect, the rle f diagnsis and f a patient s health state after a treatment shuld be determined. In additin, the beginning and the end f a treatment shuld be established. (Fr mre details, see e.g. Cutler et al. (1998) and Triplett (1999).) In 2001, Eurstat published a Handbk n Price and Vlume measurement which prvides guidelines fr the develpment f price and vlume measures fr nearly each CPA grup (Classificatin f Prducts by Activity) (Eurstat, 2001). The methds described in the Handbk were adpted in a Eurpean Regulatin issued in 2002 (Cuncil Regulatin, 2002). The bjective f the Regulatin is t harmnise the methds f deflatin used in the Natinal Accunts f Eurpean cuntries in rder t imprve the cmparability f macrecnmic statistics. The Handbk suggests sme feasible methds fr the abvementined measurement prblems. The Regulatin requires that 1 Crrespndence can be sent t: Antni Chessa, Statistics Netherlands (CBS), P.O. Bx 4000, 2270 JM Vrburg, the Netherlands. Tel.: ; acsa@cbs.nl 2 The number f emplyed persns in the health and welfare sectr was 894,000 in 1998 and 1,032,000 in 2004, which is equivalent with 13.6% f the ttal number f emplyed persns in 1998 and 14.7% in (Surce: CBS StatLine) 2

3 each Eurpean cuntry shuld use at least a s-called B methd in the Natinal Accunts starting frm Kleima et al. (2004) presented a vlume index based n hspital discharges in the Dutch Hspital Discharge Register (HDR). The aim f that study was t explre the pssibilities fr cnstructing an utput index that meets the requirements fr a B methd. We will summarise this methd in Sectin 2. In Sectin 3, we will investigate pssibilities f including services frm ther health sectrs, such as nursing hmes, psychiatric institutes and care f disabled patients. In Sectin 4, we will use the utput index fr hspital health services in rder t derive the develpment f labur prductivity fr general hspitals. We will use data n labur input and wages fr different labur categries in rder t develp an verall labur input index. The yearly develpments f health utput, labur input and labur prductivity f general hspitals will be calculated and presented fr the perid In Sectin 5, we will discuss the present state f ur research within the cntext f the requirements f the Eurpean Regulatin and the applicability f the utput methd within the Natinal Accunts. We will als discuss the pssible implicatins f uncertainties in utput and labur input data n labur prductivity. 2. An utput vlume index fr hspital health services 2.1 An utput index fr clinical and day treatments In this sectin, we summarise the methd fr calculating an utput index fr clinical and day treatments based n data frm the Hspital Discharge Register (HDR), which is a database managed by the Dutch rganisatin Prismant. Mre details abut the methd can be fund in Kleima et al. (2004). Individual inpatient treatments are recrded in the HDR. The fllwing data f discharged patients were used fr cnstructing the utput index: Date f birth, which allws us t create age classes; Diagnsis, accrding t 11,182 ICD cdes; The number f hspitalisatin days, which reflects the duratin f hspital stay; Clinical and day treatments are distinguished. Clinical treatment lasts at least 24 hurs, while a day treatment lasts shrter than 24 hurs. In ur apprach, each discharge is cunted as a treatment. (A discharge als applies t day treatments here.) The discharges are aggregated int distinct grups in rder t calculate vlume indices in a meaningful way. 4 Fr this purpse, the diagnses were characterised by the 3-digit ICD-9 classificatin, which resulted in apprximately 1000 diagnsis grups. 5 Since age and hspitalisatin duratin are nt independent fr mst diagnses, discharges were als subdivided accrding t age. Data analyses shwed that it is useful t grup discharges accrding t 7 age classes (0, 1-14, 15-44, 45-59, 60-69, 70-79, 80 and lder). The individual treatments in the HDR are thus gruped int abut 7000 diagnsis/age grups. In the Netherlands, a Cst f Diseases (CD) study is perfrmed nce every five years. Hwever, the prices that are prvided by this study are nt suitable as weights fr the diagnsis/age grups fr reasns described in Kleima et al. (2004). The main reasn is that the CD study is perfrmed nce every five years and is published abut tw years after the end f the year f review. This implies that the study des nt prvide current prices and des nt allw readjustment f the 3 A distinctin is made between A methds ( mst apprpriate methds ), B methds ( which can be used in case an A methd cannt be applied ) and C methds, which shall nt be used. 4 This was dne in rder t avid vlumes equal t zer in the reference perid, in which case vlume indices cannt be calculated. 5 Accrding t the Eurpean Cuncil Regulatin (2002), methds which use the Internatinal Classificatin f Diseases (ICD) t classify discharges can als be a B methd prvided the diagnses are recrded at a very detailed level and apprpriate cst weights are used. 3

4 weights n a yearly basis. This is a prblem because a Laspeyres index with annually adjusted weights is preferably used fr utput vlume in the Natinal Accunts. The prices per treatment that were derived in the CD study appeared t be determined fr abut 85% by the price f hspitalisatin. Analyses shwed that weighting with values frm the CD study and weighting with the number f hspitalisatin days gave similar vlume indices. 6 These findings led us t adpt the number f hspitalisatin days as weights fr the diagnsis/age grups, which are available in the HDR every year. We decided t assign the weight f ne clinical hspitalisatin day t a day treatment since infrmatin is lacking abut prices f medical peratins. The yearly vlume changes in utput are btained by calculating a vlume index fr all the diagnsis/age grups, which are weighted by the fractin f the hspitalisatin days in each grup n the ttal number f hspitalisatin days ver all the grups in the preceding year. This gives rise t a Laspeyres index, which can be expressed as fllws fr year t with respect t year t 1: T D t (1) Y t, t 1 v. T = i D In this expressin, T t dentes the number f treatments r discharges fr diagnsis/age grup i in year t and v t 1 dentes the fractin f hspitalisatin days fr diagnsis/age grup i in year t 1, which takes values between 0 and 1. The cmplete set f diagnsis/age grups is dented by D. In Sectin 4, chained indices will be derived frm (1) fr the Dutch general hspitals ver the perid (see Table 1). The utput methd described abve adds clinical and day treatments within every diagnsis/age grup. This can be mtivated as fllws. The HDR-data shw that there is a tendency fr substituting clinical treatments by day treatments (see als Table 1). Fr example, appendicitis is treated mre ften with laparscpy than with traditinal surgery. We intend t measure such a shift as a price change and nt as a vlume change. Ntice that this shift may affect the weights v t 1 f the diagnsis/age grups in (1), which are related t prices. It is wrthwhile nting the implicatins f this apprach fr medical treatments that cnsist f a series f hspital admissins. Each admissin is cunted as a separate treatment in ur methd. In case f chrnic diseases like varicse veins, it can be justified t cunt each discharge as a separate treatment, because the time between subsequent treatments will vary and the disease cannt be really cured. Hwever, in case f chemtherapy a series f admissins crrespnds with the treatment f ne diagnsis. 7 These examples shw that it is difficult t find a unifrm apprach that is crrect fr all types f diseases. 2.2 An verall utput index The HDR cvers nly clinical and day treatments. Other surveys prvide data n the fllwing additinal hspital services, which need nt be registered in the HDR: Outpatient services; Other health services, which cnsist f: Part-time psychiatric treatments; Rehabilitatin; Outpatient births; Haemdialysis; Thrmbsis services. 6 In the case where the price per hspitalisatin day is the same fr all treatments, bth weights are identical as the prices cancel ut in the relative cst weights. 7 This wuld mean that different admissins shuld be cunted as ne treatment. Ntice, hwever, that this des nt necessarily imply significant differences with ur methd. If the average number f admissins per chemtherapy is the same in tw successive years, then the vlume index f the crrespnding diagnsis/age grup remains the same. 4

5 Outpatient services are quantified as the number f visits, which in the available data are nt specified with respect t type f specialist. As a cnsequence, we cnstruct a vlume index based n the ttal number f visits in tw successive years. The vlumes f the remaining five health services are als quantified as the number f treatments r services. Psychiatric treatments and rehabilitatin refer t day treatments. The vlume f thrmbsis services is measured as the number f bld-takings per year. We derived an verall Laspeyres index by weighting the Laspeyres indices f the different types f health services. The weights were derived frm data n health insurance claims in a pilt study. These data apply t ne specific year; we assume that the weights have the same value every year. We dente the set f health services cnsisting f utpatient services and ther health services by R. The insurance claim-based weights are dented by c D fr clinical and day treatments and by c i, where i R, fr the remaining types f health services. A weight c i was derived fr utpatient visits and fr each f the five ther health services listed abve. Fr every health service type i R, we dente by T t the number f treatments, visits, r therwise, in year t. The Laspeyres index (1) fr clinical and day treatments has been extended t the fllwing verall Laspeyres index Y t,t 1 : T D t (2) Y t, t 1 = cdyt, + ci, T where the nnnegative weights c D and c i, fr all i R, are nrmalised such that they sum t 1. Output indices fr the different services are presented in Sectin 4.1 fr the Dutch general hspitals, which will be used next t derive labur prductivity. i R 3. Pssibilities f including ther health sectrs Beside hspital services, a large part f the health sectr cnsists f services prvided by: Nursing hmes; Hmes fr the elderly; Hme health care institutes; Psychiatric institutes; Institutes fr disabled patients. As is the case fr hspital services, health services frm each f these five types f institutes are prvided n the basis f budgets. The csts f health services in the Netherlands are cntrlled by a law n tariffs f health services (the WTG-law ). Health institutes submit their set f prductin agreements t the Dutch rganisatin CTG, which executes the WTG-law and derives a budget frm the prductin agreements. The ttal budget fr the five sectrs abve was almst 17 billin Eurs in 2003, which is abut 3.5% f the Grss Dmestic Prduct f the Netherlands. This budget is clse t the sum f the budgets f general, academic and specialised hspitals and the self-emplyed health prfessinals. The five sectrs thus frm a significant part f the utput f the ttal health sectr. Until nw, the utput vlume index numbers fr nursing hmes and hmes fr the elderly have been calculated accrding t a C methd in the Dutch Natinal Accunts. The methd used s far cnsists f deflating the ttal prductin value r budget by the wage-related part f the tariff f a nursing day. The purpse f this sectin is t make a brief investigatin f the pssibilities fr develping at least a B methd fr each f the five health service types. We will d this by studying the prductin data fr these sectrs. The CTG maintains a database with prductin agreements fr every year, which are specified by institute and prduct type, with a tariff-based value fr every prduct. In glbal terms, the five health sectrs treated in this sectin are characterised by the fllwing prduct cmpsitin. The main health utput f nursing hmes cnsists f nursing, which is expressed as the number f nursing days. It makes up mre than 70% f the ttal budget f the nursing sectr. Other utputs are intramural and extramural health services, such as shrt stays and supplementary 5

6 care (intramural), husehld assistance, hme care and assistance f elderly patients in their daily activities (extramural). Hmes fr the elderly are characterised by a similar prduct cmpsitin, with days f care as the main utput. Hme health care cnsists f extramural services. Health services prvided by nursing hmes, hmes fr the elderly and hme care institutes shw changes in prduct cmpsitin ver time. The number f extramural services has increased fr each f these three sectrs, in particular since 2003, with new prducts being added in each successive year. In the case f nursing hmes, the budget fr extramural services increased frm 112 millin Eurs in 2003 t 319 millin Eurs in The yearly changes in prduct cmpsitin ffer a great challenge in the develpment f an utput vlume index fr these sectrs. The prduct cmpsitin f services by institutes fr disabled patients and by psychiatric institutes shw nly slight variatins ver time. The main health utput cnsists f treatment days, which are distinguished accrding t type f handicap fr disabled patients (mentally, physically, auditry). Treatments f mentally disabled patients are differentiated further accrding t severity f handicap, patient s age and institute capacity. Treatments in psychiatric institutes are als differentiated at tw levels f detail: by grup (addicted patients, children, adults, elderly) and by intensity f treatment r assistance. The develpment f an utput vlume index fr the five sectrs is hampered by the additin f new and changing prducts in successive years. The extent f this prblem is mst serius fr nursing hmes, hmes fr the elderly and hme care, especially frm 2003 nwards. Until then, the prduct cmpsitin f these sectrs underwent small yearly changes. Frtunately, the main utput f each f the five sectrs can be tracked frm year t year, which makes it pssible t develp an utput index fr these prducts. The detailed budget and prductin data shuld enable the develpment f a B methd fr every sectr. The treatment f new prducts and the integratin f their utput vlumes with the utput indices f the main prducts is an pen questin at this stage. 4. Applicatin t labur prductivity f general hspitals 4.1 Output vlume In this sectin, we derive the verall utput index fr health services f general hspitals in the Netherlands during The purpse f this sectin is bth t illustrate the utput index methd described in Sectin 2 and t derive labur prductivity because f the attentin this measure has received in Dutch plitics and the media. We exclude academic hspitals since educatin frms a substantial part f their utput. The part f labur input vlume that is invlved with educatin is difficult t determine because separate data n this aspect are nt available. This represents a surce f uncertainty; we decided t limit uncertainties as much as pssible in this study. Table 1. Output indices fr three health service grups and verall utput index, fr all Dutch general hspitals (1998 = 100). The weights are derived frm insurance claims. The hspitalisatin days cmprise bth clinical and day treatments. Clinical and day treatments Outpatient Other services Overall index Hspital. days Discharges HDR-index visits ,46 100,55 100,31 100,19 100,51 100, ,85 101,12 100,28 100,43 100,66 100, ,97 104,58 103,02 101,25 107,67 102, ,31 112,22 109,44 103,85 115,84 108, ,18 119,03 116,02 105,30 113,20 113, ,51 126,68 123,11 111,17 118,78 119,71 weights 0,703 0,269 0,028 Surces: Statistics Netherlands, Prismant. 6

7 Table 1 shws the vlume indices fr clinical and day treatments (HDR-index), utpatient visits and the remaining five health services described in Sectin 2.2, fr which we calculated ne vlume index. Table 1 als shws the verall utput index, which is calculated frm expressin (2). The weights f the health service types in this expressin are als shwn. The indices in Table 1 are all chained indices. 8 The HDR-index is a chained vlume index that is derived frm expressin (1). Ntice that the number f hspitalisatin days decreases, while the number f discharges increases. The HDRdata shw a shift frm clinical treatments t day treatments. The verall index in Table 1 hardly increases until 2000, while it increases rapidly frm 2002 nwards. The behaviur in the first three years f the series is almst the same fr the three health service types. The subsequent utput increase cincides with changes in plicy by the Dutch Ministry f Health, Welfare and Sprt in rder t reduce waiting times. Until 2001, health institutins culd nly deliver services that were in accrdance with a fixed set f price and vlume agreements. In 2001 additinal financial means were made available by the Ministry, which allwed health institutins t make additinal agreements. Institutins culd therefre generate mre utput than n the basis f the initial set f vlume agreements. 4.2 Labur input Labur prductivity relates ttal utput t labur input. In rder t calculate labur prductivity accurately, it is f crucial imprtance t identify the characteristics f labur input that may influence utput. The OECD prductivity manual gives a cmprehensive discussin n measurement issues cncerning labur input (OECD, 2001, Chapter 4). Factrs that shuld be taken int accunt are the cmpsitin f the labur frce and its vlumetric unit. The first factr refers t the differentiatin f labur input by categry. There may be differences in skills, educatin and experience that give rise t different cntributins in utput. One hur wrked by a specialist is nt the same as an hur wrked by a member f the nursing staff. The mst apprpriate measure fr the vlume f labur input is the number f hurs actually wrked (OECD, 2001, p. 39). The number f emplyee jbs, fr instance, des nt reflect changes in the average wrk time per emplyee. Factrs like sick leave and hliday leave affect utput and shuld be included in a measure fr labur input. The data n labur input, which are presented belw, are cllected in a survey cnducted by the Dutch rganisatin Prismant. The survey cvers all the general hspitals in the Netherlands. Labur input is measured in full-time equivalents (fte s). The labur agreements n cntractual hurs f wrk during a wrking week have nt changed since 1998, s that the cnversin frm fte s t cntractual hurs has n effect n labur prductivity fr the perid cnsidered in this paper (1 wrking week is equivalent with 36 hurs). We have subdivided the hspital staff int categries based n the unifrm classificatin standard utlined by Prismant. The fllwing categries have been created: Persnnel under cntract in general hspitals: Nursing staff; Medical staff (includes specialists and ther scientific staff); Other clinical staff (labratry, psychscial supprt and ther supprt staff); Trainees; Administrative persnnel; Husing staff; Infrastructure management staff; Persnnel nt under cntract in general hspitals: Self-emplyed specialists; Other staff, mainly recruits frm agencies. 8 In all the tables and figures we set the index fr the reference year 1998 at 100, which is in accrdance with usual cnventins. Ntice that we use the value 1 in all the frmulas instead. 7

8 The grup f self-emplyed specialists cnsists f specialists wh have a practice in a general hspital. The secnd grup f persnnel that is nt under cntract in a general hspital cmprises cnsultants, dctrs and persnnel invlved in administrative functins. The develpment f fte s within each f the abve categries during is shwn in Table 2. In ur calculatin f an verall vlume index fr labur input we have adjusted the fte s by annual rates fr sick and maternity leave. The tw rates are cmbined int a single rate fr every year, which is als shwn in Table 2. The annual rates are values averaged ver all the labur categries. As we d nt have the rates at categry level, we assume that the annual rates fr sick and maternity leave are the same fr each categry. Fr every year, we calculated a Laspeyres index fr labur input by weighting the indices f every labur categry accrding t the salaries in each categry, which include cntributins t scial security payments and supplements regarding vacatin and health insurance. The weights are equal t the share f the summed salaries within each categry n the ttal sum f the salaries ver all the categries. We dente the amunt f fte s in categry i fr year t by F t and the cmbined rate fr sick and maternity leave in year t by r t. The Laspeyres labur index fr year t with respect t year t 1 then becmes: 1 r F t t (3) L t, = w, 1 r F where the salary-based weights fr labur categry i are dented by w t 1, which refer t year t 1. Of curse, the summatin in (3) is ver all the labur categries. i Table 2. Labur vlume indices fr labur categries and verall index, fr all Dutch general hspitals (1998 = 100). Administr. Husing Infrastruct. Trainees Clinical staff Nt under cntract Sick and Overall mat. leave index Nursing Medical Other Self-empld Other % ,65 99,36 98,42 96,06 100,34 110,55 101,58 99,21 111,53 7.5% 101, ,77 99,72 97,37 95,80 99,66 109,97 102,20 100,95 137,62 8.0% 102, ,43 97,80 93,68 91,94 100,91 115,25 103,86 106,64 164,99 7.7% 106, ,32 102,06 97,89 103,81 103,37 127,96 107,91 109,75 164,53 6.9% 111, ,03 100,92 98,42 100,25 104,82 136,72 111,81 115,65 132,91 6.3% 114, ,68 96,38 98,42 96,77 106,16 150,30 115,21 115,77 115,84 6.1% 115,67 weights '98 0,185 0,126 0,022 0,031 0,259 0,055 0,132 0,138 0,052 weights '04 0,146 0,074 0,020 0,024 0,319 0,096 0,177 0,101 0,043 Surces: Statistics Netherlands, Prismant; Vernet (sick and maternity leave nly). Table 2 shws mderate t strng increases within all clinical staff categries, including the selfemplyed persnnel, and the administrative staff. The fte s f medical specialists increase faster than the fte s f the nursing persnnel. This culd be related t the shift frm clinical treatments t day treatments. Table 2 als shws that recruited persnnel increases rapidly until 2001, while it decreases fast during the next years. The initial increase may be a cnsequence f labur market measures by the Dutch Ministry in rder t reduce waiting times in hspitals. The subsequent decrease f fte s amng recruits culd be the result f a mre cst-effective emplyment plicy by hspitals. (Beside the salaries f recruits, health institutes als have t pay recruitment agencies fr their services.) Ntice that the decrease in fte s, tgether with the decrease and levelling ff in ther labur categries, slwed dwn the verall labur input cnsiderably in Labur prductivity Labur prductivity is defined as the rati f utput vlume t labur input. Labur prductivity reflects hw efficiently labur is cmbined with ther factrs, such as capital, t generate utput. In this study, we decided t chse grss utput instead f value-added, since the cntributin f 8

9 intermediate labur inputs (mainly recruits) t utput is unknwn. We therefre included intermediate labur inputs in the calculatin f the verall labur vlume index. The chained index fr labur prductivity is shwn in Figure 1. Labur prductivity decreases until 2001, which is a cnsequence f the fact that verall utput increases at a lwer rate than verall labur input. In this perid, health services were delivered n the basis f a fixed set f prductin agreements. The crrespnding yearly utput vlumes did nt change much. In the last tw years we see the ppsite behaviur, in which verall utput increased at an almst cnstant high rate, while labur input increased at a lwer rate. The increase f labur input has slwed dwn, which cntributed t an increase f labur prductivity in 2004 by 4.8% with respect t Figure 1. Develpment f verall utput, labur input and labur prductivity (1998 = 100). The dashed lines are merely drawn t guide the eye Overall utput Overall labur input Labur prductivity Discussin In this paper, we described the utput vlume index fr hspital health services develped at Statistics Netherlands. We als cnsidered the pssibility f develping an utput index fr ther health sectrs, such as nursing and hmes fr the elderly. We are cnfident abut the pssibility f develping a B methd fr the ther sectrs, because very detailed utput and budget data are available fr every sectr. The main utput f each sectr can be tracked frm year t year, which culd serve as the cre f an verall utput index. The integratin f the utput vlumes fr new prducts is a great challenge that has t be dealt with in the future. Althugh the internatinal fcus is mainly n the develpment f price and utput indices that satisfy the requirements established in the Eurpean Regulatin f 2002, we shuld nt ignre ther prblems. One f these prblems is the treatment f uncertainties in utput, labur input and prductivity. In ur example f labur prductivity f general hspitals in Sectin 4, uncertainties can be identified in the cst weights f hspital health services, as these are based n an utdated surce. Variatins f 5% in the values f the weights in Table 1 affect the Laspeyres labur prductivity indices by 0.2% at mst. When academic hspital services are included, the uncertainty in the labur input share that is invlved with educatin has t be dealt with as well. The cmbinatin f different surces f uncertainty may lead t larger variatins arund the base results. The identificatin and quantificatin f uncertainty and its effect n utput, labur input and prductivity is a subject that shuld merit mre attentin. Labur prductivity is als receiving a lt f attentin in ther sectrs beside hspital health services, such as services by nursing hmes. As sn as a suitable utput vlume index has been develped fr the sectrs treated in Sectin 3, we will als quantify labur prductivity fr these sectrs. Eventually, we will have a cmprehensive picture f utput and labur prductivity fr the 9 As a cmparisn: labur prductivity f the Dutch market sectr increased with 4.5% in (Surce: CBS StatLine) 9

10 health sectr and their develpment in time, which we can cmpare between different health sectrs and als with the Dutch ecnmy as a whle. Acknwledgments The authrs wuld like t thank Dr M. Okkerse-Ruitenberg and Dr J. de Haan (Statistics Netherlands) fr their useful cmments n an earlier versin f the manuscript. References Cuncil Regulatin (EC) N 2223/96 and crrespnding cmmissin decisin N 2002/990/EC cncerning the principles f measuring prices and vlumes in Natinal Accunts (2002). Cutler, D.M., McClellan, M., Newhuse, J.P. and Remler, D. (1998). Are medical prices declining? Evidence frm heart attack treatments. Quarterly Jurnal f Ecnmics, 113 (4), Eurstat (2001). Handbk n price and vlume measures in natinal accunts, Luxemburg. Kleima, F.J., Warns, P. and Opperdes, E. (2004). Cnstructing a vlume index fr hspital services in the Netherlands. Statistical Jurnal f the United Natins ECE, 21, OECD, OECD Prductivity manual: A guide t the measurement f industry-level and aggregate prductivity grwth. ( Triplett, J.E. (1999). Measuring the Prices f Medical Treatments. Brkings Institutin Press, Washingtn, DC. 10