The referral mechanism has been introduced and made functional in most of the districts but are providing mixed results in terms of compliance and non-compliance to the defined hierarchical pattern of referral. The characteristic  of referred cases reveal various lacunae

Referral system is one of the important aspects of efficient public healthcare delivery mechanism. It defines processes for effective use of multi-tier system of health centres and hospitals for treatment of patients, according to the severity of illness. Complicated cases beyond the scope of treatment in a particular facility is stabilized first with appropriate medical care and then promptly referred and transferred to a technically equipped higher-tier hospital by following a definite referral chain.  Such a system has been introduced in most of the states of India within various administrative units.

The basic concept of a referral system is to emanate from lowest level and to end in the tertiary care facility. The referral system planning involves two decisions, which are: (i) deciding on the referral protocol, (ii) deciding on the referral chain. While referral protocols contain administrative guidelines, facility, equipment and service norms, referral chains, at present, are designed based on primarily two criteria. They are distance and availability of required service facility at the nearest point, following the hierarchical pattern of healthcare delivery.

The referral mechanism has been introduced and made functional in most of the districts but are providing mixed results in terms of compliance and non-compliance to the defined hierarchical pattern of referral. The characteristic  of referred cases reveal various lacunae. Patients often do not reach the appropriate health facilities, patients most often do not reach health facilities in time, and patients often do not follow the referral advice. Moreover, patients bypass lower level health facilities, unnecessarily overcrowding the higher  level hospitals.

The outcome of this system is mixed in terms of compliance and non-compliance to the defined hierarchical pattern of referral, due to a number of factors ranging from awareness, spatial and supportive logistics and socio-economic conditions of the beneficiaries. This is particularly relevant in areas with a difficult terrain, where spatial factors and socio-economic conditions play a decisive role in complying with the suggested referral chains, to the pre-defined hierarchical health units. Thus planning for an adequate health system with an efficient referral mechanism, requires a combination of facility and spatial analysis to derive an optimal service delivery system and GIS could serve as a useful mechanism for decision support planning, considering the incorporation of spatial and non-spatial data in a single reference frame.

The use of GIS in health has been attempted by different agencies in India. Danida-assisted National Leprosy Eradication Programme is one of the foremost in introducing GIS in health, in the country [1]. Apart from DANLEP, many development agencies [3,4,5] and government institutions are exploring health GIS in India. Malaria Research Centre, New Delhi [2], Vector Control Research Center, Pondicherry, UNICEF, WHO for leprosy, TB, Malaria and Pulse Polio programmes, HIV/AIDS programmes in TN, Orissa and MP are few of the recommended studies.

However, all these studies aim at developing health / disease maps to aid in facility and preventive planning. An interesting work has been carried out by LN Balaji [4] of NATMO Kolkata, using GIS to study the influence of  locational attributes on health conditions and also to determine the nature of disease diffusion across geographic regions. Some research has been attempted on creating health database, and using it as a support for health facility planning. The study by Mili Ghosh. Shantanu Lal and Dr. MS Nathawat of BIT Meshra[6] is on these lines and it provides a facility upgradation plan. So far no attempt has however been made for referral system design in India, as institutionalization of the referral mechanism is a relatively new management concept in public healthcare delivery system. A somewhat similar study in identifying referral regions based on the service population and catchment area features has been attempted by Dartmouth Atlas of Health Care in the United States.

There is no spatial component in the state health referral system. However, in the arduous terrains like Sundarbans, West Bengal, the spatial component has a major role to play while deciding for a referred health center. The major factors, in addition to distance and disease type/condition, are type of road, availability of river route, seasonal dependency, time of the day (day or night for river route), available conveyance type, etc.

Here we show how GIS can be used as a useful tool for decision support planning, considering the incorporation of spatial and non-spatial data in a single reference frame. Sunderban region of India, located in the state of West Bengal, has been chosen as a case study area. The present state of health referral system is devoid of specific spatial considerations, except for crude nearness estimate between the source and the destination health centers. While this absence of detailed spatial considerations may be acceptable for urban, semi-urban or even mainland rural areas that enjoy a good connectivity by rail & road; it is a cause of grave concern for arduous terrains like the Sunderbans where free movement between a source point to a destination health center ften gets heavily impaired due to spatial limitations.

We have proposed a network optimization model, based on several spatial as well as non-spatial factors, to minimize an integrated cost function. An optimization model, incorporating spatial and non-spatial data, has been proposed for designing an effective referral system model, specific to arduous terrains. The model has been developed considering the geographical spread and terrain characteristics, natural and climatic conditions, seasonal deviation, land use, infrastrutural and service facilities, connectivity and communication network, etc. and to identify the natural and physical conditions and factors limiting mobility.

The optimization model analyzes several routes from one health center to another center, using road, river or a combination of road and river, depending on several factors like disease condition, severity of disease, season, time, socio-economic condition, etc. Whenever there are changes in health center availability, new roads and river routes, the spatial database could easily be updated and new routes will be derived.

Insight into the Optimization Model

The study thus attempts to formulate a health-system-aware and terrain-sensitive referral strategy that would take the spatially dominant factors into due consideration. The overall strategy is formulated as an optimization problem where, initially, every health center is considered an equally potential referral candidate for any patient, originating from any village in the region and having any possible complaint or condition. A set of feasibility constraints is overlaid on the whole set to prune out a smaller subset that qualifies as one of the viable referral points. Finally, a composite cost function that computes the economic, temporal, qualitative and other variant costs, makes a choice of the ‘ideal’ referral point that minimizes the cost and therefore, maximizes benefits. In order keep the option open for subjective judgments that may not be captured in the model (due to lack of data and / or timely update),
we generate two or more ranked referral-point candidates and allow for a final human selection.

The Cost Function

The objective of the referral system design activity is to create a networked optimization model based on several relevant spatial as well as non-spatial factors that would minimize the cost functions, which hinder the effective use of the referral chain. Amongst various parameters, the cost function would attempt to optimize the following:

Commutation Cost

This model describes the distance between any two points, the different modes of transport used and the total cost to reach the destination. For Sundarbans, in order to reach a particular point from a given point, one has to go by the land or by the river or both. Thus the distance information is broken up as land distance and river distance respectively. The sum of the above two distances gives the total distance to travel. Likewise the total cost to travel is the sum of the costs to travel on land and the costs to travel by river.

Based on the total time (TT) and total cost (TC), commutation details defines a priority index called Accessibility Index. Comparing the accessibility indices for the different routes from one point to other, one can identify the best possible route in terms of time and cost.

Service Availability Constraint

The service model describes the services rendered by the different health centers. It identifies the name of the health center, its type (i.e., PHC, BPHC, RH, SDH, etc.), diseases/ ailments that are treated and the criticality level of the disease that can be handled. For a given patient, with a certain level of criticality and originating from a particular location, one can determine the possible destination health service centers from the service model.

Distress Factor

Distress factor is a quantitative measure of the amount of distress or discomfort that one has to bear in order to travel from one point to other. The distress factor for any two points is defined by the condition of the roads, the time in waiting for the availability of transport (worst case consideration) and the number of transport changes that one has to undergo. In case of Sundarbans all of these three factors are again dependent on the season (navigable waterways) and the time of the day (occurrence of tides).

Thus the Accessibility Index between any two points computed in the commutation model varies inversely with the distress factor between them, and combining this distress factor with the commutation details, one can redefine Accessibility Index as Qualitative Accessibility Index.

Disease Constraint

The Disease Constraint model defines all the factors (both clinical and spatial) that must be met for treating the diseases with different level of criticality. This includes the allowable time (maximum) necessary to get a particular treatment, the condition of the road required to transfer the patient to the health center, and the maximum number of transport changes or relocations that can be allowed.

The Referral (Computation) Model

The model is thematically multi-sliced with a combination of spatial and non-spatial slices intertwined on hierarchical information architecture. These slices are mostly conceptual. In terms of an implementation under a GIS system, multiple slices are flattened into a few GIS layers for efficiency of storage, visualization and computation.

At the base slice there is a spatial layer. This is where the whole story starts and this is where the story should end as well. This layer has a set of node points nodes representing the villages (origins for patient) as entry points and also the nodes representing the health centers (destinations for referrals) as exit points. Besides the entry and exit points, this layer also marks the village clusters as GPs and represents the distinguishability or indistinguishability of every village, in terms of resolution for the referral.

The next slice is again a spatial layer that represents the routing information. It comprises node points (called Intermediate Points or IP) that are used for transiting from one mode of conveyance to another or from one route to the next. A typical IP comprises of bus stops, ferry ghat, villages and health centers. Besides IPs, this layer also has the road and river network for easy computation of connectivity and routing information. This connectivity information is annotated with day/night and seasonal information that affects the routes.

Overlaid on the same slice and registered with the connectivity is the information on various routes based on different modes of transport bus, van, boat. etc. This also provides the necessary commutation data (including fare and time of travel) required in the cost function. Next couple of slices mostly maintains various non-spatial data that primarily act as constraints for feasibility of using a HC as a referral point for a patient-disease-condition point. The non-spatial information includes:

1. Available facility at HC the man-machine-medicine trio;
2. facility required to handle a disease /and or condition;
3. list of high occurrence diseases and conditions;
4. treatment conditions required for handling a disease or condition.

Process of Optimization / Constraint Satisfaction

We use iterative constraint satisfaction to refine from a global set of HCs, with the repeated application of one or more sets of constraints. The idea is to first ensure that a HC under consideration must satisfy the basic requirements for service delivery for any case under consideration. Once the constraint satisfaction is achieved, we can have one of several situations:

1. No HC is left out in the viable satisfied set. We then have no solution. This should be rare. But as and when it happens, it would be a grave warning to the health system because it indicates that the system is unable to provide any treatment path of the village/ disease combine.

2. Only one HC is left in the viable set. We know the unique referral and we are done.

3. More than one HC is retained. We then identify the cost components for optimization and compute the overall referral cost for each of the HCs. The one with the lowest cost is marked as the referral in this case.

For the case 3, we may optionally accept more than one best (least cost) solution. The final referral can then be selected based on human judgment using criteria that may not have been captured / modeled above.

Conclusion

It is indeed difficult to address the problems of public healthcare utilisation in the Sunderbans in totality, as it is extremely dependent on geophysical and natural conditions. Serious attempts are being taken to tackle some of the problems by providing supportive logistic facilities, but a planning process in-building the conditioning factors is bound to strike the problem at the root and thus create a situation for improved utilisation through appropriately designed referral chains.

Study of the referral and referred cases would reveal the degree of compliance with and utilisation of the referral mechanism and resulting improvement in morbidity status. The realistic referral chains may also give indications for facility allocation among different units, depending on the service load.

Acknowledgements

The authors duly acknowledge the financial assistance provided by the Department of Science and Technology, Government of India for carrying out the present work.

References
[1] https://www.danlep.org/gismis.html.

[2] Srivastava, Aruna and B.N. Nagpal.  Mapping malaria. GIS Dev., 4(6):28-31.

[3] Dhiman, R.C., R. Sudarshana, V.P. Sharma, M.K. Das and S.K. Bhan. Targetting mosquitogenic conditions with emphasis on Anopheles sundaicus on Car Nicobar using remote sensing and Geographic Information System techniques: A pilot study. Asian-Pacific Remote Sensing and GIS J., 13: 23-28.

[4] https://www.gisdevelopment.net/application/health/overview/index.htm

[5] https://www.gisdevelopment.net/application/health/overview/healtho0003.htm

[6] Spatial Decision Support System Using GIS based Infrastructure: Planning in Health Education for Ranchi District, Mili Ghosh,Shantanu Lal,Dr. M. S. Nathawat, Map India 2002.

[7] Referral System Planning using GIS, Chandreyee Das and Jibanananda Roy, ICMIT 2005, Kharagpur, India.

The Area of Study

The Sundarbans, located in the eastern part of India and in southern West Bengal, with a population of more than 3.5 million, spreading over 19 blocks of both the districts of 24 Parganas, is one of the underdeveloped regions in the state with predominance of small and marginal farmers.

The 54 islands, interspersed with bodies of water, are covered with forests and swarms. Wide tidal rivers and estuaries and narrow tidal creeks intersect them.  Transport and commu-nication networks are inadequate in this hostile geographical and topographical location and people have to travel in an assortment of improvised country boats, cycle-rickshaws and buses to reach their destination, which is extremely time and cost inefficient.

There are no major hospitals in the region and travel time varies between 6 to 8 hours for reaching sub-divisional or district hospitals from the core of Sunderbans. 11 RHs, 8 BPHCs  and 45 PHCs are located in the region with 659 SCs. Most of the BPHCs, PHCs and SCs are situated in the riverine area whereas RHs are located at the entry/exit point of the mainland area of Sundarbans.


The map of Sunderbans

A study on a representative sample of gynecological and obstetric cases in three selected blocks of South 24 Parganas in Sunderbans revealed that in case of patients originating from hospitals, 45 percent of the cases have complied with the designated referral chain while 55 percent did not comply. Any pre-defined norm based system of health delivery like the referral system, which may be applicable to different districts of West Bengal, may not be applicable in the case of Sunderbans, particularly because of its geographical features.

Thus planning for an adequate health system with an efficient referral mechanism calls for the design of an interactive and dynamic system for optimal design of referral chains, considering the spatial and non-spatial attributes.

It requires a combination of facility analysis along with spatial analysis to arrive at an optimal service delivery system, and GIS is the most usefu tool for decision support planning considering the incorporation of spatial and non-spatial data in a single reference frame. 

Be a part of Elets Collaborative Initiatives. Join Us for Upcoming Events and explore business opportunities. Like us on Facebook , connect with us on LinkedIn and follow us on Twitter , Instagram.

Related March 2007


whatsapp--v1