DISTRICT BASED NORMATIVE COSTING MODEL Oxford Policy Management, University Gadjah Mada and GTZ Team 17 th April 2009
Contents Contents... 1 1 Introduction... 2 2 Part A: Need and Demand... 3 2.1 Epidemiology and health seeking... 3 2.2 District context and Demand modelling... 4 2.3 Demand modelling... 5 2.4 Non-SPM utilisation... 6 3 Part B: Supply... 6 3.1 SPM direct resource sheets... 6 3.2 Overheads... 7 4 Model output for scenarios... 9 References... 10 1
1 Introduction The objective of the district based normative model is to provide realistic, dynamic costing of the SPM (Standard Pelayanan Minimal) services that are part of the government s commitment to health of the population of Indonesia. The model is designed to be used to simulate the costs of providing services in a district based on key characteristics of that district. As such it takes account of the structure of health services in the district, geographic location and specific disease characteristics of the area. Whilst the objective of the model is to cost the SPM it is clear that services that are not part of the SPM make a significant and probably increasing contribution to the workload of facilities. Since many resources within the system are shared between services (SPM and non-spm) the apportionment of these overheads to SPM services is partly dependent on the provision of non-spm services. Furthermore facilities have a maximum workload beyond which they cease to be economically and clinically effective requiring further capital investment, again a function of both SPM and non-spm activity. For both these reasons it was felt important to ensure that that some measure of non SPM activity was included for the workload modelling even though the model does not provide a detailed cost of these activities. Figure 1: Main menu The model is divided into two main parts. The first part Demand/Need provide an estimation of workload either based on need for services or demand. The second part, Supply, provides information on the direct and overhead costs of providing service. Combining the cost components with workload provide a series of simulation of costs both for a chosen scenario year and over time. All parts of the model can be accessed from a main menu (Figure 1). 2
2 Part A: Need and Demand 2.1 Epidemiology and health seeking District-wide workload is based on assumptions about the incidence, prevalence and need for services entered on the Incidence sheet. For each of the conditions information is entered on the target group for service (Column E - pregnant women, births, under fives etc) and the prevalence of the disease (Column F - expected number of cases for each person in the target group). Using information on the population structure of a given district this provides an estimate of the number of cases of the condition in the district. Whilst for some conditions all cases are assumed to need medical treatment (at a health facility or through outreach care by medically trained personnel) for other conditions self-treatment is an option for many perhaps most cases. This is the case, for example, for most cases of diarrhoea where only a small proportion are expected to require medical intervention. The proportion expected to require treatment by medically trained personnel is entered in column G. Detailed SPM targets are entered in column G. This is the proportion of need (Target Population x Prevalence x Those requiring treatment) that is expected to be covered under the SPM by the target year. The maximum for all services is 100% but less than 100% might be considered on the grounds of affordability or community willingness to use services (an alternative more complex modelling of demand is also available and considered in section 2.3). The final two data entry columns are those requiring hospital care and those self-referring. The first of these (Column H) is a normative based on the division between those that can be safely and effectively treated at the Primary (Puskesmas and below) level and those that require hospital based treatment. The second column (Column I) reflects actual practice of patient self-referring themselves to hospital for primary treatment. The patient journey is simplified in the model as follows (patient distribution is show on the sheet Condition Need ): Figure 2: Patient journey through the costing model 1-F Self- Treatment J Puskesmas Inpatients Target Group E Episodes of Illness Episodes requiring treatment E Condition Prevalence (%) F Requiring treatment (%) G Covered by SPM (%) H Requiring Hospital Treatment (%) (referrals) F G Covered by SPM Puskesmas Patients Hospital Patients Puskesmas Outpatients Hospital Outpatients I Direct Referral (%) J Admission Rate (%) K Patients admitted to hospital from sub-districts without beds (Number) (referrals) Hospital 1-J Inpatients 1-I I H 1-H 1-J J K 3
Some of the patients treated at the Puskesmas but not requiring hospital level care (not referred or self-referred) may still require admission. For those sub-districts with puskesmas equipped with beds these are assumed to be treated within the sub-district. Those without beds are transferred to the district hospital. Transport for self-referrals are assumed to be financed by the patient. For patients presenting at the puskesmas but requiring onward referral to the district hospital either because the require hospital level care or because there are no beds in the puskesmas are assumed to be financed by the health system. The cost of referral is based on the distance between the originating puskesmas and the district hospital. 2.2 District context and Demand modelling The model can be used to simulate the costs a range of different district contexts. District specific information is entered on three different sheets. These are: 1. District information on condition prevalence (sheet = Incidence ) whilst it is expected that much of this data will not vary substantially across districts prevalence of some conditions may need to be locally determined. In particular the prevalence of some communicable diseases such as dengue and malaria will vary substantially and can be entered individually. The default is that the district data take the national figures. 2. Demographic information on the district is entered on the sheet District profile. This includes information on each sub-district including population, the type of sub-district facility available and number of beds. Also the demographic structure of the population for the district as a whole by target groups (number or proportion of births, pregnant women, women of reproductive age etc). Also the per capita household income in Rupiah and population growth rate. 3. Geographic data entry (Sheet = Map ) which shows the location of each puskesmas (Marked 1 to 40), private facility (marked P1, P2 etc) and district hospital (marked H1, H2 etc). East- West and North-South scales can be altered to accommodate districts of different sizes. Rough sub-district boundaries are drawn in using the e key. The model then computes the size of each sub-district 1 and the estimated distance travelled by those living in each subdistrict to the sub-district puskesmas, district hospital and all other puskesmas in the district. This information is used in two ways in the model. First it is used to estimate the costs of referral by puskesmas to the district hospital. Second, it is used in the demand equations to estimate demand for services when demand based scenarios are specified. Scenario data can be saved or reloaded from the main menu (select Save/Load District Scenario ). 1 This makes use of an approximation algorithm that divides each sub-district into a series of eight non equal triangles radiating out from the puskesmas centre the border of each sub-district. A macro then calculates the area of each triangle and sums them to provide an estimate of area. Simulations suggest that even with extremely irregular sub-district areas the overall error in estimating areas is less than five percent. 4
2.3 Demand modelling The model produces two types of estimate of cost. The first are based on need modelling where utilisation of services is estimated as a proportion of total need as estimated using epidemiological data. This estimate ignores the other factors that influence whether people seek services from a particular provider in the event of illness. As an alternative, and following Bitran (Bitran 1991), the model will predict demand for SPM services based on a three stage demand function. This is based on a three level nested logit model that first determines the probability of seeking care, second whether care is sought from a puskesmas, private primary provider or hospital and finally which provider (see Figure 3). Figure 3: Structure of the modelled demand for SPM services Illness Care seeking (Prseek) Care seeking Choice of level (Prpusk) Hospital Puskesmas /private PHC Self-treatment /no care Choice of provider (Prchoose) Hospital 1 Hospital 2 PHC 1 PHC 2 PHC 3 Probabilities are specified as follows: 5
Where u is the indirect utility associated with no care, M are PHC facilities (puskesmas or private) and N are hospitals. The coefficients, and are respectively one minus the correlation coefficients of the primary care provider options, hospital provider options and choice between hospital and primary care. V is the indirect utility associated with a particular choice of provider and is specified as followed (Bitran 1991): where is the price a user pays for service at provider j, is per capita income in sub-district s 2, is the average weighted distance from people living in sub-district s to provider j, incorporates the effect of other variables affecting utility of utilising services (implicitly assumed to be constant for households across the district 3 ), is the coefficient on price (<0), the income coefficient (>0) and the distance coefficient (<0). Demand by users in all (S) sub-districts is computed for each provider. This leads to a (M+N).S demands across all providers to given total demand for each provider. The main challenge is in calibrating the model to provide a realistic estimate of demand. This is best done by examining the price, income and distance elasticities and ensuring that these match estimates from regional or country-based empirical (econometric work). 2.4 Non-SPM utilisation The final component of the need/demand specification are provided estimates for non-spm utilisation. Although no direct estimate of the cost of non-spm services is provided, non-spm workload is required in order to properly apportion the overhead costs between SPM and non-spm services. The following information is required 1. Estimate of direct staffing time required for an average non-spm outpatient visit and inpatient day at both the puskesmas and hospital levels. 2. Non-SPM outpatient visits per 1000 population for puskesmas and hospital. 3. Non-SPM inpatient episodes and average length of stay per 1000 population for puskesmas and hospital. 3 Part B: Supply 3.1 SPM direct resource sheets Personal SPM services were divided into four main groups - Maternal Health, Child Health, Other Reproductive and Communicable and 33 conditions (see Annex 1). Differences in 2 Note that we simplify this in the model assuming that income is constant across the district given the lack of sub-district level income data. 3 This is a strong assumption. In particular it is likely that variables such as educational levels, which will vary across the district, will affect the demand for services. 6
treatment in puskesmas network and the District Hospital meant that for most conditions separate sheets were required for each condition. The approach adopted focus group interviews with experts who have local experience in the management and treatment of the specific conditions identified to obtain estimates of resource use for each SPM. A total of 60 SPM resource sheets captured the information on personal SPM services. Population based SPM services such as disease control measures (e.g. spraying for dengue) as well as overhead costs such as training and monitoring and evaluation were captured separately (see SPM overhead). A resource sheet, adapted from the Core Plus Costing Model developed by MSH, was used to assemble information on the resources required to treat each condition. These sheets are also similar in appearance to other ingredient based costing tools as WHO s Mother and Baby Package. Each sheet collects information on: staffing requirements - minutes of time required by activity for one episode of the condition) drugs and medical supplies supply required by dosage, numbers of times per day, days of treatment and expected proportion of patients requiring treatment laboratory tests and radiology examinations type of test/examination, quantity and proportion of cases requiring procedure Each sheet also identifies the average number of visits required for each episode, the inpatient admission rate and average length of stay should admission be required. These variables are used for computing the proportion of overhead allocated to each service. The information provided on the SPM sheets are summarised on two sheets: Medical supplies summarises the costs of drugs, supplies and tests for each of the 60 condition sheets together with information on overhead requirements (visits and admissions). Staffing summary summarises the minutes of time required of direct staffing by condition Each of these sheets are then multiplied by district-wide workload to derive the total direct resource requirements for each condition. 3.2 Overheads Overheads in the model are divided into three types: 1) facility based overhead covering the costs of running hospitals and health centres including annualised; 2) DHO overheads specific to individual SPMs; and 3) DHO administration. 1. Facility Overhead This covers the costs of running the three types of facility included in the model - puskesmas, puskesmas with beds and district hospitals. Items covered are annualised capital costs of the building and equipment, maintenance, utilities (fuel, electricity etc) and administrative supplies. These are entered in aggregate on the sheet Facility overhead. Administrative and service staff are treated equivalently but separately in the model. 7
Overheads can vary in a number of ways according to numbers of staff, workload (adjusted beddays), population of the sub-district or numbers of facilities. The model user is required to specify the drivers of the main overhead categories (sheet = Overhead assumptions ). More than one driver can be specified by allocating percentages. So, for example, equipment particularly that utilised in providing direct care, might be partly thought to be determined by workload, and partly fixed to the number of facilities. The default driver is the number of facilities. That proportion of overhead that is assumed to vary in relation to the number of facilities are initially fixed. Larger workloads may, however, lead to a need for an expanded number or size of facilities. This signalled in the scenarios and automatically increased. 2. SPM specific Much of DHO spending is on items that are specific to particular SPM services. These are classified into 11 categories 1. Registration and case finding 2. Training 3. Recording and reporting 4. Monitoring and Evaluation 5. Materials 6. Specimen collection and laboratory testing 7. Health Promotion 8. Surveillance/analysis of data 9. Coordination 10. Services 11. Equipment Each resource is allocated to an SPM group and/or condition (sheet SPM specific overheads ). The proportion allocated to population prevention as opposed to personal care for conditions is also specified (the default is personal care). This allocation makes no difference to the total spending by SPM condition. Episode specific costs, however, only include the proportion allocated for personal care. Without such a provision all the costs of prevention would be allocated to individual episodes of treatment. So, for example, the costs of treating dengue would include the substantial costs of preventing dengue for the entire population leading to a gross distortion of the unit costs of care. 3. DHO overheads This includes those costs that are treated as overhead to entire district system to be allocated across SPM and non-spm services and then between SPM services (sheet DHO Costs ). These include the staffing costs of administering the districts plus other district-wide costs. Adjusted bed-days are used as the way of allocating these costs. 8
4 Model output for scenarios The supply and demand parts of the model provide the main building blocks for costing scenarios. Once these are specified, costs can be projected based on a number of core parameters (assumptions). These are specified from the Summary Results page ( Results Summary ) as follows: Target year year by which the targets specified should be achieved Scenario year the year for which detailed costs are required (year by year costs until the target year can also be retrieved by going to Annual Costs (sheet = Annual cost ). Demand/need basis specify either scenarios based on need or demand. Capacity assumptions specify whether the model should incorporate estimates for expanding the facility variant fixed costs (this provides a more realistic estimate of the cost of expanding capacity) Assumptions about the achievement in the base year (Base SPM) SPM Targets - by default these are based on the detailed condition specific targets specified on the Epidemiology sheet. It is also possible to over-ride this by specifying general targets (10-100%) for all SPMs. Workload assumptions a. the proportion of time staff spend on direct service provision relative to time spent on administration or waiting to provide care (including on-call time) b. the number of hours that direct staff spend on publicly financed activities (hours can be varied from 1 to 10 hours a day). This control has been included given the concern than many staff spend a relatively small proportion of their time on public activities. The substantial impact on cost of varying these hours is immediately apparent. The cost estimates on this page provide an immediate summary for the scenario year. Please note that these costs are automatically adjusted for year specific factors such as population growth and inflation (see general assumptions). More detailed results are provided on additional output pages including: 1. SPM Costs total and per capita costs by SPM condition broken down by type of cost (sheet = SPM Costs ) 2. Sub-District Costs - total and per capita SPM costs by sub-district. These are based on even distribution of disease burden across sub-districts (sheet = Sub district Costs ) 3. Annual Costs Successive recalculations are undertaken to provide an year by year estimate of SPM costs by condition from the base to target year (sheet = Annual cost ). These are based on a gradual achievement of targets (detailed on sheet = SPM Targets ) In addition to the annual point estimates of cost it is possible to undertake basic sensitivity analysis. The user clicks on the Sensitivity Analysis box to be presented with a series of variables that can be used varied. The user has the option either of specifying a lower and upper value for the variable or specifying a mean, standard deviation and confidence level (95% is the default) and lower and upper values that match the confidence interval are generated automatically. Only total costs are generated for all values between the lower and upper value for the base to target year (sheet = Sensitivity analysis ). 9
References Bitran, R. A. (1991). "Health Care Demand in Developing Countries: A Model of Household Demand and a Market Simulation Model of Health Care Financing." Boston University, Ph.D. 10
Annex 1: Main groups and SPM conditions Main Group Condition Group 1 Maternal Health Basic Antenatal Care Pregnant Women 2 Maternal Health Abortion Pregnant Women 3 Maternal Health Antepartum Haemorrhage Pregnant Women 4 Maternal Health Hypertension PET Pregnant Women 5 Maternal Health Severe Anaemia Pregnant Women 6 Maternal Health Premature Labour Births 7 Maternal Health Abnormal fetal presentation Births 8 Maternal Health Prolonged Labour Births 9 Maternal Health Caesarean Section Births 10 Maternal Health Uterine Rupture & Hysterectomy Births 11 Maternal Health Intrapartum & post partum Births infection 12 Maternal Health Post Partum Haemorrhage Births 13 Maternal Health Normal Delivery Births 14 Maternal Health Routine Post Partum Care Births 15 Maternal Health Neonatal Complications Births 16 Child Health Routine Infant Health Infants <1 17 Child Health Routine Child Health Children 1-5 18 Child Health Child Immunisation Infants <1 19 Child Health Nutrition for the Poor 6-24 months 20 Child Health Severe Malnutrition Children <5 21 Child Health School Health Children 7 years old 22 Other Reproductive Family Planning Non Permanent ELCOs Health 23 Other Reproductive Family Planning Permanent ELCOs Health 24 Communicable Pneumonia Children <5 25 Communicable Diarrhoea < 5 years Children <5 26 Communicable Diarrhoea > 5 years Children & Adults >5 27 Communicable Malaria < 5 years Children <5 28 Communicable Malaria > 5 years Children & Adults >5 29 Communicable Tuberculosis <5 years Children <5 30 Communicable Tuberculosis >5 years Children & Adults >5 31 Communicable Dengue < 5 years Children <5 32 Communicable Dengue > 5 years Children & Adults >5 11
Main Group Condition Group 33 Communicable 34 Public Health Acute Flacid Paralysis Population 12