|
nif:isString
|
-
Punjab is the 15th most populous state in India, with 22 districts. It has an overall population of 27 million, 13.7% growth rate and 73% literacy rate [13]. The state MMR of 155 (95% confidence interval; 85,226) per 100,000 live births is below the national average of 178 [7]. ERS service was launched in Punjab on 3rd April, 2011 under public-private partnership with public financing and private delivery. A state-level 24×7 call centre and office of the private provider was established at Amritsar district. Aim of ERS was to provide free transport service to pregnant women, neonates, post natal cases, infants and children with ill-health, victims of road side accident and for all other health emergencies in the general population. ERS in Punjab was implemented in four different phases. Out of 240 ambulances, 90 ambulance were launched in the first phase (April 2011) which were stationed in 12 districts, 50 ambulances in 18 districts in second phase (July 2011) and another 50 ambulances serving all 20 districts in third phase (August and September 2011). Finally, 50 more ambulances were added in fourth phase (October and November 2011). Distribution of vehicles was based on a criterion of 10 ambulances per million population. Out of the 22 districts in Punjab, 2 districts were created out of existing districts after the introduction of ERS system in Punjab. These two new districts were part of their parent districts in the Management Information System (MIS) of the service provider. Hence we report results for 20 districts. The ambulances were stationed in the district at district hospital, community health centres and primary health centres. Ambulance could be called to place of emergency by dialling a toll-free number ‘108’ in any district throughout the state. Nearest ambulance available is dispatched by call centre operator using a well-functioning geographic information system (GIS). ERS is primarily limited for transportation to public sector institutions except if there is strong insistence by beneficiary/attendants for private facility.
Data Collection: Source and Methods: Both primary and secondary data was collected. Secondary data for almost 0.4 million calls received during one year period from 1st April, 2012 to 31st March, 2013 for all the 20 districts of Punjab, as recorded in database of 108 service provider was obtained. This data was drawn from the ERS tracking software which is generated from the calls received at centralized call-centre for availing ERS. The response to the call, i.e. whether service was availed or declined was also fed into the software by call centre operator. The information captured in this data source comprised of time taken to respond to emergency, type of health care provider, i.e. public or private, level of service provider etc. Month-wise and district-wise data on public sector institutional deliveries in Punjab from the year 2008 to 2013 was obtained from Punjab Heath System Corporation (PHSC). Primary data on users and non-users of 108 referral service was collected at facility level in three districts i.e. Amritsar, Roopnagar and Sangrur of Punjab. Users were those patients who reported in the health facility, having utilized 108 ERS service; while ‘non-users’ were those who reached the health facility by any other means of transport. Choice of selection of these three districts was primarily based on the performance of ERS, which was determined by number of calls per ambulance per day or calls per ambulance per million population. As there was not much diversity geographically therefore the level of infrastructure in the district was also kept under the view while selecting the districts. Sangrur had lowest rate of utilization, Roopnagar - medium utilization, and Amritsar was randomly chosen from among group of high performing districts. Health facilities within the district were selected randomly. In the health facility, consecutive sampling was used for selection of users and non-users. All patients who reported during the period of data collection were included in the study. Patients were interviewed using pre-tested semi-structured to collect data on basic socio-demographic characteristics, reasons for using and not using 108, client satisfaction, level of severity etc. among users and non-users. We recruited a total of 411 users and 999 non-users of 108 referral transport were recruited in the present study. The quality of 108 ambulance service in Punjab was evaluated using a checklist designed by Delhi Government as a standard for a Basic Life support (BLS) ambulance [14]. All the ambulances in 21 randomly selected health facilities in Amritsar and Sangrur district were inspected against this checklist to assess their adequacy in terms of infrastructure, design, availability of drugs, consumables and other life support systems besides trained technicians and manpower. We estimated the economic cost of implementing referral service in Punjab from the health system's perspective. Cost data for the 108 ERS was collected from the accounts department of the PHSC. Non-recurrent or capital costs and recurrent or operational costs were elicited. The non-recurrent costs comprises of cost of 240 vehicles, set up cost of emergency response centre in district Amritsar, information-technology infrastructure (including software) and pre-operational costs i.e. trainings, recruitments, marketing, administration and communication costs. Operational costs were the annual costs paid to private provider for delivering the service, which was fixed per ambulance, subject to a minimum of 3 calls per day per ambulance.
Data of 391369 calls, received from 1st April, 2012 to 31st March, 2013 was analyzed to assess the extent and pattern of utilization. Utilization pattern was assessed by the gender, medical condition of the patient and type of health facility. To ascertain the impact of 108 referral transport service on equitable utilization of public health facilities, we analysed the primary data on users and non-users of 108 using multiple logistic regression.
To evaluate the quality of the referral service in Punjab, the primary data on users and non-users of 108 along with the data on physical inspection of 21 ambulances was analyzed. We used propensity score matched (PSM) non-users for each user (Table 1). Propensity score is the conditional probability of an individual to utilize 108 referral service from the group of users and non-users individuals taken together in the presence of covariates [15]–[17].
Table data removed from full text. Table identifier and caption: 10.1371/journal.pone.0109911.t001 Overview of Study Methodology for Evaluation of Emergency Referral Services in Punjab State, India. The regression analysis showed that the users and non-users of ERS were significantly different on a set of independent factors. Any comparison of time taken to reach the health facility from the time of emergency could thus be confounded by these factors. Other studies also show that these factors influence the different types of delays in accessing health facility for obstetric care [18], [19]. We used propensity score matching to generate a probability score for each individual (user or non-user) for using ERS, contingent upon a set of characteristics. The covariates which were included for generating propensity score included distance of place of emergency from health facility, locality of residence of the individual, education of head of household, socio-economic status of household, and district. Nearest neighbourhood method within permissible calliper of 0.02 (7% of the mean propensity score) was used. Overall the matched sample comprised of 276 pairs of users and non-users. Mean time taken to reach the health facility among users and matched non-users was computed. We analysed the effectiveness of ERS on reducing the 2nd delay of the 3-delay model of Kwast, which means the time taken since the decision is made to seek medical care till the pregnant woman reaches health facility [20]–[23]. Additionally, client satisfaction among 108 users and proportion of calls attended for users by the time of the day (derived from calls data) were computed as quality measures of the referral transport system. Physical quality of ambulance was measured against BLS benchmarks (Table 1).
Impact of Referral Service on Public Sector Institutional Delivery: Trends in public institution deliveries before and after the introduction of 108 referral service were assessed (Table 1). It was assumed that intervention would start displaying its effect after the first phase itself when 90 ambulances (out of 240 planned) were already active for service delivery. Interrupted time series analysis was undertaken using Segmented Linear Regression (SLR) [24]. SLR works on the equation given below; Presence of autocorrelation was examined upto 12th lag using Durbin Watson test statistics and autocorrelation and partial autocorrelation functions [25]. The coefficients for secular trend (β1), change in the level of public institutional deliveries after the intervention was introduced (β2) and trend of the public institutional delivery in post-intervention period (β3) were estimated along with their 95% confidence intervals. Best fitting linear model was obtained after adjusting for autocorrelation and seasonal variation [24]. A sensitivity analysis was done with varying assumptions of time of treatment effect i.e. after first, second, third and fourth phase of implementation. Four prediction models were tested which differed based on the assumption of showing treatment effect at different time points. In the base model, it was assumed that the intervention shows its effect after first phase (April 2011). In the 2nd and 3rd model, it was assume that the intervention shows its effect after second phase (July 2011) and fourth phase (November 2011) respectively. In the final model, it was assumed that the intervention shows its effect 6 months after the complete implementation of the service. The crude birth rate per 1000 population declined from 17.6 in 2007 to 16.2 in 2011 [26]–[30]. This could have potentially confounded the analysis. In order to test this, we used ‘proportion of deliveries which occurred in public sector institutions’ (out of total pregnancies) instead of total monthly public sector institutional deliveries as the dependent variable in SLR. Using the annual figures on births, population and growth rate, month-wise estimations for total population and births in Punjab was done using best fitting exponential curve. Numbers of pregnancies were estimated using an assumption of 10% pregnancy wastage. Finally, the proportion of pregnant women delivering in public institutions of Punjab was estimated using public institutions delivery data and total estimated pregnancies. We examined the dose-response relationship by estimating correlation between the “proportion of deliveries where 108 service was used” in a district (indicative of ERS performance) with the district level “beta coefficient of impact of intervention” (indicative of impact of ERS on institutional delivery). Costing was done from a health system perspective. The analysis of cost incurred on the referral service was based on broad classification i.e. capital and recurrent costs (Table 1). We annualised the costs of all the capital items based on the average life span of each item and discounting the costs at 3%. Equivalent uniform annual costs for each capital item were computed. We used cost paid to private provider per ambulance during 2012 as operational cost. Unit cost was estimated as cost per call, cost per individual transported and cost per kilometre travelled.
The study was approved by the Institute Ethics Committee of the Post Graduate Institute of Medical Education and Research, Chandigarh, India. Administrative approvals were obtained from the Health Department of Punjab state. Written approval was obtained from Civil Surgeons of the concerned districts and officer in-charge of health facilities which were visited. Written informed consent was also obtained from all study subjects (patients and staff) who were interviewed.
|
![[This page as RDF]](http://data.gesis.org/somesci/static/rdf-icon.gif){kind=icon}