Study setting
CHSI is the first national costing study commissioned by the Department of Health Research, Government of India to generate empirical evidence on the cost of health care delivery in secondary and tertiary level health facilities [
26]. The CHSI analysis specifically focussed on estimating the cost of actual resources spent by the health system in the provision of health services.
Under CHSI, a total of 38 public facilities, comprising 11 tertiary care and 27 district hospitals, and 16 private hospitals were included in the sample from 11 states of India. Out of these 11 states, four were selected from the north region (Jammu & Kashmir, New Delhi, Rajasthan and Uttar Pradesh), three from the east region (Bihar, West Bengal and Odisha), two each from the west (Gujarat and Maharashtra) and south region (Andhra Pradesh and Tamil Nadu). In addition to geographical representation, these states were chosen to represent the variation in net state domestic product (NSDP), health indicators and health workforce density across India. `The multi-stage sampling strategy also aimed to capture differences in cost associated with specialties and type of providers. The procedure followed for selection of each of the public (district and tertiary facilities) and private hospitals are explained in detail in the protocol paper [
26].
A mixed methodology consisting of both bottom-up and top-down costing approaches were used for data collection, and standard analytical principles were applied [
27,
28] The lack of disaggregated data on resource use and electronic health records in the Indian healthcare system led to the use of mixed costing approach. The unit cost of outpatient consultations, inpatient bed days and intensive care bed days were estimated using the top-down approach. The cost of an individual surgery was estimated using a mixed micro-costing approach. Under this, the data on the use of resources like equipment, drugs, and consumables for each surgery was captured using a bottom-up approach and the cost of human resources, infrastructure, furniture, and overheads was estimated using the top-down methods. Combining both the approaches provides with a sufficient degree of disaggregation of the estimated cost into its specific input resources, necessary for the purpose of setting reimbursement rates and HTA. The data collection was undertaken for the reference period of April 2017 to March 2018 across all the sampled hospitals. The details on the data collection methodology and data analysis plan of the CHSI study and a process evaluation of the quality and challenges faced during the data collection have also been published elsewhere [
12,
26,
29].
Analytical approach
We estimated speciality specific unit cost of services within each selected hospital. These unit costs included the cost-of-service delivery in four basic cost centres of outpatient department (OP), inpatient department (IP), intensive care unit (ICU) and operation theatre (OT) within a speciality. The cost-of-service delivery under each of the cost centres were computed following a standard classification of fixed and variable costs. Cost of input resources that are not dependent upon on the output produced, i.e., salaries of human resources, annualized cost of capital space, equipment (excluding the maintenance cost) and furniture were classified under the category of fixed costs. Further, the costs, which vary with the increase or decrease in the volume of output, e.g. drugs, consumables, utility, stationery, maintenance other supplies and overheads such as electricity, water, maintenance, etc., were classified as variable costs.
The unit costs of service delivery for each centre were computed based on the actual resource consumption and service utilization (i.e., current levels of capacity utilization) of the health facilities. However, as service utilisation (e.g., outpatient consultation, number of inpatient admissions both in inpatient wards and ICU) relative to the resources available (i.e., capacity utilisation) varies across similar services, specialities and facilities, unit costs were standardised to enable comparison. As bed occupancy rate is a standard indicator reflective of hospital service load, it was used to adjust for differences in capacity utilization for each of OP, IP, ICU and OT specific standardized unit costs [
10,
30,
31]. Standardised unit costs were calculated using the service utilisation figures (the denominator in the unit cost) in line with bed occupancy rates of 80% and 100% of full capacity for each speciality. Bed occupancy rates were calculated based on actual data on the number of beds, average length of stay and patients admitted during the particular year. Under the standardization process, the cost incurred for variable resources such as drugs, consumables, utility, overheads, etc. were adjusted for the change in capacity utilisation while keeping cost of fixed assets in the form of space, equipment, furniture, and human resources constant. All costs were analysed in Indian Rupees, 2020 prices and converted to USD for presentation (1 USD = ₹ 76.21) [
32].
A total of 327 specialties were included, with 48, 79 and 200 specialties covered in tertiary, private and district hospitals respectively (Table
1). Further, from these specialties, cost data collected from a total of 408 OP units, 327 IP units, 45 ICU units and 219 OT units were included in the analysis (Supplementary tables S
1 –
3). Distribution of the cost centres by the type of provider, location of the facility (by the tier of the facility) and by specific facilities is presented in Table
1 and supplementary tables S
1 –
3.
Table 1
Profile of the sampled specialities at each facility by type of facility and tier city
Inpatient Cost Centre |
Overall | 327 | 22.0 | (10.0-45.0) | 3.9 | (2.8-5.1) | 0.7 | (0.3-1.4) |
By type of facility | p < 0.05 | p < 0.05 | p < 0.05 |
District | 200 | 27.0 | (13.8-48.0) | 4.3 | (3.3-5.2) | 0.8 | (0.4-1.6) |
Private | 79 | 6.0 | (3.0-11.0) | 2.5 | (2.0-3.0) | 0.5 | (0.3-0.9) |
Tertiary | 48 | 52.0 | (32.3-18.3) | 5.8 | (4.0-7.0) | 0.7 | (0.5-1.2) |
By tier city | p < 0.05 | p < 0.05 | p < 0.05 |
Tier1 | 25 | 5.0 | (4.3-5.5) | 42.0 | (30.0-63.0) | 0.5 | (0.3-0.7) |
Tier2 | 81 | 3.0 | (2.0-4.7) | 11.0 | (5.0-30.0) | 0.5 | (0.3-1.1) |
Tier3 | 221 | 4.0 | (3.0-5.2) | 24.0 | (10.0-45.0) | 0.8 | (0.4-1.6) |
ICU Cost Centre |
Overall | 45 | 14.0 | (10.0-24.0) | 3.5 | (2.0-5.0) | 0.6 | (0.3-1.5) |
By type of facility | p < 0.05 | p < 0.05 | p < 0.05 |
District | 19 | 14.0 | (10.0-22.0) | 4.6 | (4.1-5.7) | 0.7 | (0.4-1.6) |
Private | 10 | 10.5 | (9.3-14.0) | 3.0 | (2.9-3.0) | 0.2 | (0.1-0.4) |
Tertiary | 16 | 18.0 | (12.8-25.3) | 2.0 | (2.0-3.3) | 0.8 | (0.4-1.6) |
By tier city | ns | p < 0.05 | p < 0.05 |
Tier1 | 9 | 13.0 | (12.0-24.0) | 2.0 | (2.0-2.0) | 0.7 | (0.4-1.5) |
Tier2 | 11 | 13.0 | (9.5-17.0) | 3.0 | (2.0-4.0) | 0.6 | (0.2-2.0) |
Tier3 | 25 | 18.0 | (10.0-29.0) | 4.5 | (3.0-5.7) | 0.6 | (0.3-1.5) |
A descriptive-analytical approach was used to present and summarise the cost data and to compare the influence of provider type on the unit costs at the specialty level for each cost centre. The role of capacity utilisation (bed occupancy) in driving the differences in unit costs across provider types was explored by comparing provider capacity utilisation unadjusted and adjusted costs. Next, the impact of the average length of stay (ALOS) on unit cost was examined by comparing the adjusted and unadjusted costs per admission and costs per bed day across the provider types. Finally, the impact of geography and price were explored by comparing the capacity utilisation adjusted costs per outpatient visit and cost per bed day across city tiers. The analysis presents the median unit costs and tests for differences using the Kruskal–Wallis test for small samples [
33]. Unit costs also vary with scale of activity, as a result of economies of scale, in a non-linear fashion, to form a classic “u-shaped” average cost curve [
34]. Where average costs are minimised, relative to the scale of activity, services are said to be scale efficient [
35]. Scale efficiency was explored by testing for the likelihood of a non-linear relationship between scale and unit cost using Pearson’s rank correlation. In addition, scatter plots with LOWESS smoothing were generated to allow the visual assessment of the relationship. Lowess smoothing is a process built into statistical software that creates a line through the central tendency of the relationship between two variables [
36]. Due to the need for large samples and to ensure comparability of service provision, the scale analysis was carried out for district hospitals only (
n = 278). The analysis was also restricted to the inpatient and outpatient cost centres as the ICU sample was relatively small and small scale, while for the OT cost centre variable costs (costs that vary directly with the level of output/scale) are a significant proportion of costs so that economies of scale are unlikely. Scale variables used were number of visits, number of admissions, number of beds and bed occupancy. Analysis was carried out using RStudio [
36].