Locational analysis of public and private health services in Rohtak and Bhiwani districts, (India), 1981 to 1996 A location-allocation problem involves three basic elements: (a) a set of consumers (demand) distributed spatially over an area, (b) a set of facilities (service centres) to serve them (Taylor, 1977) and (c) network data connecting demand points to service centres. Our main objective is to determine population weighted optimal locations, which can be solved by the following mathematical formulation (ReVelle and Swain, 1970):  i = demand location J = candidate facility location N = number of demand locations M = number of candidate facility locations P = number of facilities to locate wI = weight at demand node i dij = shortest distance between demand location i and candidate j yj = 1, if facility is located at site j; 0 otherwise xij = 1, if demand location i is served by facility at site j, 0, otherwise "i = demand site decision variable Location-allocation models were run separately for each of the services under study. For each service a model is run in four steps. In the first step candidates and demand items are defined. Each candidate (village in this study) can have either of three values: 0,1,2 indicating 'village is not a candidate', 'a mobile candidate (which means a service may be or may not be located)' and 'a fixed location' respectively. In the second step, a model criterion (such as minimum distance or maximum coverage) is defined. The 'minimum distance' criterion is employed in the present analysis, as our major thrust is on the identification of geographical efficiency of existing services in terms of weighted distance. In the third step, the system performs the locational analysis based on input parameters set in the first two steps. Solving a location-allocation problem would require analysing every possible alternative configuration and it could be a large number. Therefore, heuristic algorithms are needed to solve such problems. The Global Regional Interchange Algorithm (GRIA) (Murray and Church, 1994) and the Teitz and Bart Heuristic (TAB) are widely used algorithms for location-allocation problems (see Kumar, 1999 and ARC/INFO, Online Help for a detailed discussion on these algorithm). In the present study, location-allocation models were run three times for each service: firstly, for the identification of actual demand-weighted distances from the service centres to the demand points. Secondly for the identification of optimal/modelled centres and their average distances from the demand features allocated to them. Finally, for the identification of additional proposed locations. While running these models for actual services, villages having a facility/service were declared as pre-defined fixed locations. Once the locations of service centres are declared fixed, the model has to select these locations and will produce statistics in relation to pre-defined locations. For the identification of modelled (near to optimal considered in our analysis) locations the same model was run, but all candidates were coded as 1 (except uninhabited villages, coded with 0). This means that all inhabited villages were candidates for the service location; and rest of the procedure remains the same. The identified-modelled locations may not necessarily be the actual locations. The results of location-allocation stored in the out files were used to map the identified centres and demand locations. Results: Availability, accessibility and efficiency level of health services: A wide range of health services is available in the selected districts, extending from an RMP doctor to a civil hospital and a sophisticated medical college. But the pattern of availability varies across these two districts, and between rural and urban areas (Figure-3). In the present analysis, only two services are examined at length, viz. public and private. The public health services include CHC and PHC. Under the basic minimum needs (BMN) programme the Government set a population criterion for the provision of public health services in rural areas: a sub-centre for every 5000 people, a PHC for every 6 sub-centres (30,000 people) and a CHC for every 4 PHC (120, 000 people). All these services are inter-linked with each other and form a hierarchy, but only PHC provides the first direct link between the people and clinical, curative and preventive health services through its sub-centres (which it operates directly) . The CHC is a second level health service, but the main objective of a CHC is to diagnose 'high risk patients', and to control and manage PHCs; and as such CHC is not a first link between people and public health services, but is rather the first referral unit. Therefore, among public health services only PHCs are included in this paper. Primary Health Centre (PHC): In the selected two districts, there were 20 PHCs in 1981 and their number increased to 89 in 1996, more than four times increase over a period of 15 years. In 1981, each PHC served a population of more than 90,000 (Table-3), declining to just more than 32,000 people (a little more than the national norm) in 1996. In 1981, PHCs or higher order services were available at 30 locations (rural and urban) and their number increased to 103 in 1996. Each PHC served 30 other villages in 1981 against only 9 villages in 1996. The actual locations of PHCs are examined in relation to the modelled locations derived from the location-allocation models (Table-4). The average population weighted distance between PHC and demand points was 7.8 kilometres in 1981, declining to 4.4 kilometres in 1996. In Bhiwani, people had to travel on average one kilometre further to access a PHC as compared to Rohtak District, showing poor geographical access to PHCs in Bhiwani compared to Rohtak District. Eastern parts and areas along the National Highways have relatively better geographical access to health services as compared to southern and southern-west parts of the study area (Figure-4). Although, the actual average weighted distance declined from 1981 to 1996, but north-south contrast in the distribution of geographical accessibility to health services remained explicit. |