Interfacility transfer of pregnant women using publicly funded emergency call centre-based ambulance services: a cross-sectional analysis of service logs from five states in India

Objective To estimate the proportion of interfacility transfers (IFTs) transported by ‘108’ ambulances and to compare the characteristics of the IFTs and non-IFTs to understand the pattern of use of ‘108’ services for pregnant women in India. Design A cross-sectional analysis of ‘108’ ambulance records from five states for the period April 2013 to March 2014. Data were obtained from the call centre database for the pregnant women, who called ‘108’. Main outcomes Proportion of all pregnancies and institutional deliveries in the population who were transported by ‘108’, both overall and for IFT. Characteristics of the women transported; obstetric emergencies, the distances travelled and the time taken for both IFT and non-IFT. Results The ‘108’ ambulances transported 6 08 559 pregnant women, of whom 34 993 were IFTs (5.8%) in the five states. We estimated that ‘108’ transferred 16.5% of all pregnancies and 20.8% of institutional deliveries. Only 1.2% of all institutional deliveries in the population were transported by ‘108’ for IFTs—lowest 0.6% in Gujarat and highest 3.0% in Himachal Pradesh. Of all ‘108’ IFTs, only 8.4% had any pregnancy complication. For all states combined, on adjusted analysis, IFTs were more likely than non-IFTs to be for older and younger women or from urban areas, and less likely to be for women from high-priority districts, from backward or scheduled castes, or women below the poverty line. Obstetric emergencies were more than twice as likely to be IFTs as pregnant women without obstetric emergencies (OR=2.18, 95% CI 2.09 to 2.27). There was considerable variation across states. Conclusion Only 6% institutional deliveries made use of the ‘108’ ambulance for IFTs in India. The vast majority did not have any complication or emergency. The ‘108’ service may need to consider strategies to prioritise the transfer of women with obstetric emergency and those requiring IFT, over uncomplicated non-IFT.


STRENGTHS & LIMITATIONS
• This study is the first to assess the role of the '108' ambulance service -the largest provider of the free emergency medical services in India -for inter-facility transfers (IFTs) of pregnant women.
• We estimated the number of institutional deliveries in five States of India and the proportion of mothers who were transported by '108' for IFTs and obstetric emergencies. This helped us assess the role of '108' ambulance services for pregnant women with emergencies.
• We assessed the proportion of obstetric emergencies and type of emergencies among pregnant women who were transported by '108' as IFTs and compared them with those who were transported by '108' as non-IFTs.
• The '108' database mostly did not record data for treatment given en-route, and doctors' notes on inter-facility transfers, thus we could not study these data.

BACKGROUND
India had an estimated 83% women delivering in health facilities in 2013. [1] Despite this high proportion of institutional births, India had an estimated maternal mortality ratio (MMR) of 167/ 100,000 live births and an early neo-natal mortality rate of 28/1000 live births in 2013. [2] The country accounts for 17% (50,000) of global maternal deaths per year [3] and 26% (696,000) of global neonatal deaths. [4] Many maternal deaths occur during transit to health facilities. [5][6][7] Inter-facility transfers for pregnant women are crucial, especially in resource poor countries where most peripheral health facilities provide only uncomplicated birthing or basic emergency obstetric care. [8,9] About 14% to 36% of women delivering in facilities are referred from lower to higher-level facilities.[10-13] As inter-facility transfer is more likely to be due to referral for high risk pregnancies or complications during pregnancy, childbirth or post-partum, it can play a pivotal role in reduction of maternal morbidity and mortality. [14,15] Ensuring an uneventful inter-facility transfer is part of good health care provision at the referring facility, and will reduce delays in access to appropriate health care (delay type-2) and delays in getting appropriate care after reaching a health facility (delay type-3). [16,17] A good process will include prompt arrangement of transport, en-route stabilisation, communication with the referral facility to prepare them for the patient, and appropriate hand-over on arrival.
Inter-facility transfer is thus a complex coordinated effort made by the referring health care provider, the en-route attendant, the receiver at the referral health facility, and the referral transport system.   1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  transport. This study was conducted to investigate inter-facility transfers of pregnant women in India using '108' ambulance services. The objectives of the research were to i) estimate proportion of all institutional deliveries in the general population that were transported by '108' and the proportion of all institutional deliveries that were transferred between facilities by '108', ii) to estimate the proportion of all transfers of pregnant women by '108' that were transfers from one health facility to other (inter-facility transfers), and iii) to compare the characteristics of the inter-facility transfers and non-inter-facility transfers, across States.

Context
The '108' ambulance service operates under a public private partnership. It operates 7361 ambulances, and transfers any medical emergency across 21 States and union territories   9 should assist the delivery en-route and transfer the mother and child to the nearest health facility.
[28] For inter-facility transfers, the EMT consults the '108' call-centre based medical officer and the referring health care provider to discuss indication of referral, stability of case to withstand travel, stabilising care required and place of referral [Source: Expert from GVK-EMRI).

Working definitions
Inter-facility transfer (IFT), for this study, was defined as any transfer of a pregnant woman from one health facility to another health facility on the advice of a health care provider, using a '108' ambulance. All other transfers of pregnant women to health facilities using '108' ambulances were defined as 'non-IFT'. These mostly included transfers from home to facility.
An obstetric emergency was defined as any life threatening medical complication in women in pregnancy, labour or childbirth, or after (within 42 days of termination of pregnancy).

Study Design
This cross-sectional study analysed '108' ambulance records from five States for the period April 2013 to March 2014. Ethical approval for the study was obtained from Indian Institute of Public Health-Hyderabad and London School of Hygiene and Tropical Medicine.

Obtaining data and data management
Formal permission to use the data was obtained from GVK-EMRI. Anonymised information on '108' calls from 1 st April 2013 to 31 st March 2014 was obtained from the GVK-EMRI emergency response centre database. Data were extracted onto Excel sheets, and converted to STATA 10.0 files. Data were inspected to assess consistency, range, and missing data. Any gross issue relating to the quality of records was noted, and records with improbable entries were excluded from analysis. Variables were recoded wherever needed. Variables of interest were: inter-facility transfer; age of pregnant woman; social caste; economic class (below or above poverty line); region (rural or urban); type of emergency; time of call; day of call; time taken by ambulance to reach the client; time taken to reach the health facility; and distance travelled.
Districts within States were stratified into high priority districts and non-high priority districts.
High priority districts are those identified by the Government of India as being in the lowest quartile of districts (or tribal districts) in each State with respect to maternal and child health indicators (including institutional delivery rate, maternal mortality and neonatal mortality rates among a total of 16 indicators).   11 and institutional deliveries in the population for the study States, as mentioned below. The numbers of pregnant women transported by '108' as recorded in the call-centre database were compared with these population estimates for each State.
For each State, the number of pregnancies in the study period was estimated as sum of estimated pregnancies in rural and urban population [population (rural) X crude birth rate (rural) X 1.1 X 1000] + [population (urban) X crude birth rate (urban) X 1.1 X 1000]. The population data were obtained from the 2011 census and the crude birth rates from the Sample Registration System 2013. The multiplier 1.1 is used to account for an estimated 10% of the pregnancies which may have ended in abortions or intra-uterine deaths.
The number of institutional deliveries in the study period was estimated as [Estimated number of pregnancies (rural) X institutional delivery rate (rural) X 100] + [Estimated number of pregnancies (urban) X institutional delivery rate (urban) X 100]; Institutional delivery rates include live births and still-births. The institutional delivery rate were obtained from DLHS-4 and AHS-2 surveys.
Information was analysed for all States combined and separately by State, comparing IFT and non-IFT. The characteristics of the women transported, distances travelled and the time taken by '108' ambulances, were described for both IFT and non-IFT journeys. The association between socio-demographic and clinical variables and the outcome (IFT vs no-IFT) was investigated using bivariate and multivariate logistic regression.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  Social caste and economic class were missing in 55% and 95% of observations in the State of Chhattisgarh, and 14% and 22% in Himachal Pradesh. Given the nature of the variables we decided that it was likely that the data were missing not at random, and therefore multiple imputation was considered inappropriate. In order to investigate possible selection bias resulting from missing data, we performed a complete case analysis for each State and all States combined, supplemented by a series of sensitivity analyses. Our first sensitivity analysis involved running a model with and without States of Chhattisgarh and Himachal Pradesh.
Presence or absence of these States did not substantially change the pattern of results. The second sensitivity analysis was performed by running a model with and without social caste and economic class for all States combined (total) and for individual States. The presence or absence of these variables did not substantially change the magnitude of the pattern of results for the total and for States (except Chhattisgarh). For our main analysis (all States combined) we thus included social caste and economic class variables and excluded any data from Chhattisgarh. Models for individual States (except Chhattisgarh) also include social caste and economic class.

Study Populations
The study States had population sizes ranging from 6 million to 60 million and had different social compositions (Table-1). The percentage of rural population ranged from 61% in Telangana to 90% in Himachal Pradesh while scheduled castes and tribes, together, ranged from 22% in Andhra Pradesh and Gujarat to 43% in Chhattisgarh. The crude birth rate was

Estimated proportions of pregnancies, institutional deliveries and IFTs transported by '108'
Table-2 presents estimates of the number of pregnant women, obstetric emergencies, institutional deliveries and inter-facility transfers in the general populations, and the proportion of these transported using '108' ambulances in the five states, and overall. The 608,559 pregnant women transferred comprised 16.5% of all estimated pregnancies, and 20.8% of all estimated institutional deliveries for the study States combined.

Characteristics of pregnant women transferred by '108' ambulances
The pregnant women who were transferred by '108' belonged mostly to lower social and economic sections-rural or tribal areas (84.6%) or scheduled castes or tribes (64.9%) and below-the-poverty-line status (76.6%). Two hundred and forty two women (4 per 100,000) died before the ambulance reached the pick-up site. The proportion of pregnant women who died before arrival of '108' ambulance was higher in Chhattisgarh (0.15%) compared to other States.

Time of day and day of week when pregnant women were transferred
The transfers by '108' were more frequent during 9:00 hours and 15:00 hours, peaking around 12:00 hours with a second smaller peak at around 22:00 hours, in all the States (Figure-1).
Telangana, Himachal Pradesh and Gujarat had slightly higher proportions of transfers on Mondays, and Chhattisgarh on Thursdays. Across all the States, it was evident that a higher proportion of IFTs compared to non-IFTs were occurring between 10:00 hours to 23:00 hours.

Determinants of Inter-facility transfer
Overall, women with obstetric emergencies transported by 108 were roughly twice as likely to have an IFT as women with no obstetric emergency (crude OR 2.25, 95% CI 2.16-2.34) ( Table-5).

DISCUSSION
This is the first study assessing inter-facility transfers for pregnant women using the '108' ambulance service-the largest provider of the emergency medical services in India. We discuss findings with respect to patterns of use and the existing health system.
We estimated that '108' transferred around one fifth of all pregnancies and institutional deliveries in the five States. However, only 1.2% of all institutional deliveries were transported by '108' for inter-facility transfers. Only 1% of all institutional deliveries were transported by '108' for obstetric emergencies. The findings suggest that the '108' service is not a preferred choice for transport to a higher-level facility, or for obstetric emergencies. The proportion of IFTs among all institutional deliveries will depend on the pattern of use of level of health care, referral practices and the availability of transport for between facility transfers.
Roughly one-half of the non-IFTs went to peripheral birthing centres or basic EmOC centres. A systematic review from India (including most studies from public health facilities) found that between 14% and 36% of all pregnancies were referred from nurse-run delivery or basic EmOC centres, and a further 2% to 7% were referred from doctor-run basic EmOC centres for complications or emergencies.
[13] Assuming the above pattern of use of health facilities we found that only 5 of the 32 mothers who were transferred between facilities used an ambulance.
[6] However, for other public transportation schemes, it was found that a high proportion (two-thirds) of all the inter-facility referrals in a study from Madhya Pradesh used Janani Express service (non-ambulance) while others used personal transport, taxis, autorikshaws or public transport. [10] Although IFTs in our study were twice as likely to transport pregnant women who had any obstetric emergency compared to non-IFTs, there was a very large proportion of IFTs with no emergency or complication (92%). One of the '108' doctors, during discussion to understand IFT processes, mentioned that on many occasions they were not convinced of the need for IFT.
However on insistence of the referring staff, the '108' doctor approved transport for IFT

37]
A lag of about one hour was observed between peak hours of call for non-IFTs and IFTs. It is plausible that women first reached a facility from home using '108' (non-IFT) and, after initial assessment, some were referred to another facility using the '108' service. The lag time was required to determine need of referral, provide initial care and arrange notes. A higher proportion of IFTs immediately after 10:00 hours could also be due to the fact that medical officers usually arrive on duty at 9:00 hours to conduct rounds, and would refer the sick cases admitted on the previous night. We had large proportions of missing information on social and economic status from two States. We considered that the missingness was not at random and was not associated with outcome. There is evidence that in such situations a complete case analysis, as we reported, is associated with negligible bias compared to a multiple imputation approach. [40]

CONCLUSION
Of all the estimated institutional deliveries in India, only a very small proportion made use of the '108' ambulance for transfer between facilities. Among '108' users for IFTs around 92% did not have any complication or emergency. After adjusting for confounding factors, IFTs were more likely for women with obstetric emergencies, more than 30 years of age and from urban areas. Pregnant women from socially disadvantaged castes, below poverty line and from high priority districts were less likely to have IFTs. Utilisation of the '108' service and its determinants varied across States. Primary research is required to understand variation in utilisation and to explore the potential of the '108' ambulance service to serve a higher proportion of women requiring IFTs, in particular those having obstetric emergencies. The '108'

ACKNOWLEDGEMENTS
We acknowledge GVK-EMRI for sharing the data and allowing scientific study without any interference.

This work was supported by a Wellcome Trust Capacity Strengthening Strategic Award to the
Public Health Foundation of India and a consortium of UK universities.

COMPETING INTERESTS
The authors have declared that no competing interests exist.

Introduction
Background/rationale 2 Explain the scientific background and rationale for the investigation being reported 6-8 Objectives 3 State specific objectives, including any prespecified hypotheses 8

Study design 4
Present key elements of study design early in the paper 9 Setting 5 Describe the setting, locations, and relevant dates, including periods of recruitment, exposure, follow-up, and data collection 9-10 Participants 6 (a) Give the eligibility criteria, and the sources and methods of selection of participants 10 • This study is the first to assess the role of the '108' ambulance service -the largest 2 provider of the free emergency medical services in India -for IFTs of pregnant women. 3 • We assessed the characteristics of pregnant women who were transported by '108' as 4 IFTs and compared them with those who were transported by '108' as non-IFTs, for five 5 state. 6 • We did extensive cleaning and management of data to drive appropriate information. 7 • '108' service did not record information on post-partum obstetric emergency separately 8 thus these could not be estimated. 9 • Diagnosis of obstetric emergency may be inaccurate and subject to inter-observer bias. 10 • The '108' database mostly did not record data for treatment given en-route, and doctors' 11 notes on IFTs, thus we could not study these data.

12
• Some population estimates are based on assumptions and may not be accurate.

13
• We had large proportions of missing information on social and economic status from two 14 states. We did a complete analysis and sensitivity analysis to deal with missingness. where GVK-EMRI operated '108' service had been fully functional for more than 3 years were 22  Social caste and economic class were missing in 55% and 95% of observations in the state of 7 Chhattisgarh, and 14% and 22% in Himachal Pradesh. Given the nature of the variables we 8 decided that it was likely that the data were missing not at random, and therefore multiple 9 imputation was considered inappropriate. In order to investigate possible selection bias 10 resulting from missing data, we performed a complete case analysis for each state and all states   Description of 108 calls relating to pregnancy 7 Across the five states, '108' call centres received 646,656 calls for pregnancy-related transfers, 8 6.2% of which were for IFT. Among IFT calls, an ambulance was not assigned to 1.2%, and an 9 ambulance was assigned but not used by 6.0% of callers. Among non-IFT callers these 10 proportions were 1.4% and 3.9% respectively. A total of 608,559 pregnant women were 11 transported using '108', and of these journeys 5.8% were for IFT.     Of the pregnant women transferred by '108' ambulances, 34,993 (5.8%) had an IFT (Table-3). 8 The proportion of IFTs among women transported by '108' was highest in Himachal Pradesh 9 (11.3%) followed by Andhra Pradesh (9.9%), Telangana (8.7%), Chhattisgarh (3.2%) and Gujarat 10 (2.4%).      (Table-4). 3 4 Distance and time travelled by 108 ambulance for pregnant women 5 The median distances travelled and time taken by '108' ambulances to transfer pregnant 6 women are shown in Table-   Determinants of Inter-facility transfer 1 Overall, women with obstetric emergencies transported by 108 were roughly twice as likely to 2 have an IFT as women with no obstetric emergency (crude OR 2.25, 95% CI 2.16-2.34) ( Table-5). 3 In the adjusted analysis (excluding Chhattisgarh), obstetric emergencies had 1.95 (95% CI, 1.83- 4 2.06) times higher odds of having IFT compared to non-emergencies (  This is the first study assessing IFTs for pregnant women using the '108' ambulance service- 5 the largest provider of the emergency medical services in India. We discuss findings with 6 respect to patterns of use and the existing health system.  facilities in our study and evidence from the review, we estimate that between 40,000 and 3 80,000 institutional deliveries who used '108' for non-IFTs may require further referral to 4 higher facility. In addition, among the estimated 80% (2,300,000) institutional deliveries who 5 went to their first facility by other means of transport, some women may be referred further. 6 Thus the absolute numbers of pregnant women referred and requiring transport for IFT are 7 likely to be large while '108' transports only about 35,000 pregnant women for IFT. 8 9 In our analysis, of all the transfers by ' Although IFTs in our study were twice as likely to transport pregnant women who had any 2 obstetric emergency compared to non-IFTs, there was a very large proportion of IFTs with no 3 obstetric emergency or complication (92%). One of the '108' doctors, during discussion to 4 understand IFT processes, mentioned that on many occasions they were not convinced of the 5 need for IFT. However on insistence of the referring staff, the '108' doctor approved transport 6 for IFT (personal communication). Often the referral was done because there was no doctor on 7 duty or other resources were not available, as was also found in Madhya Pradesh.
[10] These 8 non-emergency IFTs will add unnecessary load at higher facilities and also make the 9 ambulances unavailable for other emergencies. The median time from call to '108' and reaching the destination facility was similar for non-IFT, 21 and IFT, and it ranged from 1 hour to 1.3 hour. A study of 57 maternal death reviews from Uttar 22 The role of the '108' service in improving care at the referring and referral facility, and its 4 overall impact on maternal morbidity and mortality reduction, could not be estimated from our 5 analysis nor has been reported in other studies. We acknowledge GVK-EMRI for sharing the data and allowing scientific study without any 3 interference.    1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 Introduction Background/rationale 2 Explain the scientific background and rationale for the investigation being reported 6-8 Objectives 3 State specific objectives, including any prespecified hypotheses 8

6
• We did extensive cleaning and management of data to drive appropriate information. 7 • '108' service did not record information on post-partum obstetric emergency separately 8 thus these could not be estimated. 9 • Diagnosis of obstetric emergency may be inaccurate and subject to inter-observer bias.

10
• The '108' database mostly did not record data for treatment given en-route, and doctors' 11 notes on IFTs, thus we could not study these data.

12
• Some population estimates are based on assumptions and may not be accurate.
[10] These 8 non-emergency IFTs will add unnecessary load at higher facilities and also make the 9 ambulances unavailable for other emergencies. The median time from call to '108' and reaching the destination facility was similar for non-IFT, 21 and IFT, and it ranged from 1 hour to 1.3 hour. A study of 57 maternal death reviews from Uttar 22 Pradesh found much longer times; the mean time taken to arrange transport and travel from 1 home to facility was about 4 hours, and transport from one facility to another was about 10 2 hours.
[6] Although the '108' service has a mandate to inform the destination facility before 3 arrival to reduce delays in treatment, this is not practiced as the list of contact points is not 4 provided to EMTs.[Source: Dr. GVR Rao] Few women were transferred to a district different 5 from the originating district due to sparse distribution of referral facilities. These women also 6 travelled longer distances. Continuity of care and monitoring of IFTs can be better if transfers 7 are in same administrative unit i.e. district. 8 9 The '108' ambulances are stationed close to CHCs and should be readily available, but if the 10 ambulance is on route for pick-up or drop-off of another client then the IFT client with 11 complications will have to wait or arrange for another means of transportation. The role of the '108' service in improving care at the referring and referral facility, and its 4 overall impact on maternal morbidity and mortality reduction, could not be estimated from our 5 analysis nor has been reported in other studies. Nonetheless, the '108' service is accepted as an between type of facilities could not be assessed. There is a possibility that few calls for IFT for 5 the women residing in rural areas were wrongly recorded, as they called from the health 6 facilities in urban areas. This would have contributed to higher proportion of urban women 7 among IFTs. The '108' service from Chhattisgarh was taken over by the '102' service since 8 October 2013 thus use of '108' may be under-estimated for this state. We had large 9 proportions of missing information on social and economic status from two states. We 10 considered that the missingness was not at random and was not associated with outcome. 11 There is evidence that in such situations a complete case analysis, as we reported, is associated 12 with negligible bias compared to a multiple imputation approach. utilisation and to explore the potential of the '108' ambulance service to serve a higher 1 proportion of women requiring IFTs, in particular those having obstetric emergencies. The '108' 2 service would benefit by having a triage system to ensure that women with an obstetric 3 emergency requiring an IFT are prioritised. We acknowledge GVK-EMRI for sharing the data and allowing scientific study without any 7 interference.   1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 Introduction Background/rationale 2 Explain the scientific background and rationale for the investigation being reported 6-8 Objectives 3 State specific objectives, including any prespecified hypotheses 8