Difficulties leaving home: A cross-sectional study of delays in seeking emergency obstetric care in Herat, Afghanistan

Social Science & Medicine 73 (2011) 1003e1013

Contents lists available at ScienceDirect

Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed

Difficulties leaving home: A cross-sectional study of delays in seeking emergency obstetric care in Herat, Afghanistan Atsumi Hirose a, b, *, Matthias Borchert a, c, Homa Niksear d, Ahmad Shah Alkozai e, Jonathan Cox a, Julian Gardiner a, f, Khadija Ruina Osmani g, Véronique Filippi a a

London School of Hygiene & Tropical Medicine, UK U.S. Centers for Disease Control and Prevention, USA Institute of Tropical Medicine and International Health, Charité e Universitätsmedizin, Berlin, Germany d Herat Regional Hospital, Afghanistan e World Vision International, Afghanistan f Birkbeck College, UK g Faculty of Medicine, Herat University, Afghanistan b c

a r t i c l e i n f o

a b s t r a c t

Article history: Available online 29 July 2011

This study used an analytical cross-sectional design to identify risk factors associated with delays in care-seeking among women admitted in life-threatening conditions to a maternity hospital in Herat, Afghanistan, from February 2007 to January 2008. Disease-specific criteria of ‘near-miss’ were used to identify women in life-threatening conditions. Among 472 eligible women and their husbands, 411 paired interviews were conducted, and information on socio-demographic factors; the woman’s status and social resources; the husband’s social networks; health care accessibility and utilisation; care-seeking costs; and community characteristics were obtained. Decision and departure delays were assessed quantitatively from reported timings of symptom recognition, care-seeking decision, and departure for health facilities. Censored normal regression analyses suggest that although determinants of decision delay were influenced by the nature and symptoms of complications, uptake of antenatal care (ANC) and the birth plan reduced decision delay at the time of the obstetric emergency. Access to care and social networks reduced departure delay. Programmatic efforts may be directed towards exploiting the roles of ANC and social resources in facilitating access to emergency obstetric care. Published by Elsevier Ltd.

Keywords: Afghanistan Near-miss Maternal mortality Delays Antenatal care Birth plan Social resources Obstretrics Care-seeking

Introduction Most maternal deaths in resource-poor countries are preventable if women suffering complications during pregnancy and childbirth receive emergency obstetric care (EmOC) in a timely manner (Paxton, Maine, Freedman, Fry, & Lobis, 2005). EmOC interventions range from administration of antibiotics, oxytocic drugs, and anticonvulsant and manual procedures to blood transfusion and surgery. They are usually provided in health centres and hospitals. In 1994, Thaddeus and Maine proposed an analytical framework to identify barriers to EmOC services. Called the ‘three delays’ model, this framework has been widely adopted by the safe * Corresponding author. U.S. Centers for Disease Control and Prevention, 1600 Clifton Road, NE (MS-A06), Atlanta, GA 30329, USA. E-mail address: [email protected] (A. Hirose). 0277-9536/$ e see front matter Published by Elsevier Ltd. doi:10.1016/j.socscimed.2011.07.011

motherhood community and is frequently quoted by public health professionals despite its restricted focus on emergency curative services as opposed to primary and secondary prevention. The framework distinguishes three time periods: (1) From onset of complications to decision to seek care, (2) from decision to reaching the appropriate health facility, and (3) from arrival in facility to treatment. Factors prolonging the first period are complex and often context-specific, including a woman’s status in her family and community, income constraints, perceived high costs of services or poor quality of care, traditional beliefs, and low awareness of danger signs and symptoms of severe complications (Koblinsky et al., 2006). The second period is often prolonged by travel distance and lack of facilities and transportation means. The third period relates to the quality of health care services (Thaddeus & Maine, 1994). In recent years, the focus of the safe motherhood community has shifted from two opposing paradigms (EmOC for complications

1004

A. Hirose et al. / Social Science & Medicine 73 (2011) 1003e1013

vs. skilled birth attendance for all births) to a more comprehensive approach bringing these two strategies together (Campbell & Graham, 2006). Nonetheless, the three delays framework has continued relevance in regions where many women continue to deliver at home. Hospital audits conducted in these regions have persistently documented a considerable number of moribund women reaching facilities (‘near-miss upon arrivals’), renewing calls for research into care-seeking delays (Adisasmita, Deviany, Nandiaty, Stanton, & Ronsmans, 2008; Filippi, Richard, Lange, & Ouattara, 2009; Roost, Altamirano, Liljestrand, & Essen, 2009). While many quantitative, qualitative, or mixed-method studies have investigated care-seeking behaviours for maternity services, studies quantifying care-seeking delays for EmOC are mostly maternal deaths case series, describing types or causes of delays without an explicit comparison group (Barnes-Josiah, Myntti, & Augustin, 1998; Orji, Ogunlola, & Onwudiegwu, 2002; Supratikto, Wirth, Achadi, Cohen, & Ronsmans, 2002; Urassa, Massawe, Lindmark, & Nystrom, 1997). The few existing analytic studies explicitly comparing women with and without delays have methodological limitations, such as small sample sizes (Okonofua, Abejide, & Makanjuola, 1992; Okusanya, Okogbo, Momoh, Okogbenin, & Abebe, 2007). While descriptive case series are useful in formulating hypotheses, analytic study designs can help identify delay factors that are amenable to interventions, by comparing the distribution of the outcome value between risk and non-risk groups. Our quantitative study used such a study design to identify risk factors predisposing women to delay in Afghanistan. Afghanistan is a challenging environment for women wishing to obtain care. It was ranked 174th out of 178 countries on the Human Development Index after decades of conflicts, during which the country was persistently hindered from socioeconomic development. The Afghan health care system was almost completely destroyed when the Taliban regime ended in 2001. The population’s health status is reportedly one of the worst in the world, with the national maternal mortality ratio (MMR) at 1600e2200 per 100,000 live births (Bartlett et al., 2005). The Afghan Ministry of Health (MOH) and its international partners have started to rebuild the health system by contracting out the delivery of a ‘Basic Package of Health Services’ (BPHS) to nongovernmental organisations (NGOs) (MOH, 2003). The BPHS consisted of four levels of services: (1) Health posts providing limited care in their smaller communities; (2) outpatient Basic Health Centres serving a larger population, with a referral connection to Comprehensive Health Centres (CHCs); (3) CHCs offering a wider range of services, including basic management of obstetric complications; and (4) District Hospitals providing all services in BPHS, including EmOC. This health care provision model has quickly expanded services and improved the accessibility and quality of care to the poor, yet the provision of EmOC in district hospitals remains inadequate (Hansen et al., 2008). Despite the opening of midwifery schools across the nation, utilisation of skilled birth attendants (SBA) (i.e., a person with professionally obtained midwifery skills) is low due to widespread poverty, difficult geographical access, and strict gender rules (Mayhew et al., 2008). Increasing the accessibility and uptake of EmOC services is essential to reducing maternal deaths in Afghanistan (Chowdhury, Ahmed, Kalim, & Koblinsky, 2009; Fournier, Dumont, Tourigny, Dunkley, & Drame, 2009). Afghanistan’s social system is strongly patriarchal. Men control women’s mobility, as the tribal honour codes prescribe that men protect women’s chastity, which is tied to family honour. Kinship relations are central to Afghans’ lives. Marriage usually involves a bride price payment. The poor may exchange daughters to cancel out such payments. Marriage between cousins is common because of a reduction in payment and the familiarity of the two families involved. Marriage as a way of ending a family dispute has

reportedly decreased (Smith, 2009a). The way in which a woman’s marriage is contracted largely determines her position in the marital home (Smith, 2009b). Marital homes are considered a suitable place to give birth. Methods Study setting We conducted a cross-sectional survey between February 2007 and January 2008 of women arriving at the maternity ward of Herat Regional Hospital in life-threatening conditions. We chose Herat Hospital because it is one of the largest in the country, with 17 obstetricians and 40 beds at the time of the study. The hospital represented the only comprehensive EmOC facility in the province, and complicated cases from neighbouring rural provinces were often referred there. In 2002, the MMR in Herat province was estimated to be 593 per 100,000 live births (Amowitz, Reis, & Iacopino, 2002). Inclusion criteria This study included women of all ages in life-threatening conditions requiring immediate intervention to prevent their likely deaths (often referred to as ‘near-miss’ cases) (Say, Souza, & Pattinson, 2009). At the time of the study, the World Health Organisation (WHO) had not yet standardised near-miss criteria (Say et al., 2009). Disease-specific criteria of near-miss complications were therefore adapted from other studies conducted in resource-poor settings (Filippi et al., 2005; Prual, Bouvier-Colle, de Bernis, & Breart, 2000), which may have been less stringent than the newly established WHO criteria. Per the criteria adapted from other studies, a woman must exhibit one of the following eight conditions during pregnancy, labour, or 42 days after termination of pregnancy upon admission, irrespective of pregnancy outcome: 1. Impending rupture of uterus characterised by Bandl’s ring 2. Clinical diagnosis of rupture of uterus 3. Eclampsia characterised by convulsion with urine protein of 2 þ on a dipstick and diastolic blood pressure (BP) > ¼ 90 mm Hg, or convulsion or coma with or without high BP in the absence of other causes 4. Severe pre-eclampsia with diastolic BP >110, urine protein of 3 þ or more on a dipstick, and two additional symptoms of preeclampsia 5. Vaginal, intra-abdominal, and concealed bleeding with an episode of shock, or requiring an IV therapy of 2000 cc or more fluids given through two or more IV lines or an emergency hysterectomy 6. Severe infection characterised by abdominal pain with temperature >38  C or <36  C not explained by an extragenital infection plus two signs of severe infection, or an episode of shock. 7. Severe maternal anaemia (haemoglobin level < 7 g/dl) with dyspnoea and requiring transfusion of two or more units of blood 8. Acute heart failure requiring intravenous furosemide From the above eight criteria, ten complication types were created, by dividing the fifth criterion, haemorrhage, into three types according to the gestational age at admission. Unmarried women were excluded because we knew from experience that interviews would cause them emotional distress as well as disturbance in the ward; out of wedlock pregnancies are typically considered taboo in Afghan culture.

A. Hirose et al. / Social Science & Medicine 73 (2011) 1003e1013

Case identification and data collection All women arriving in the maternity ward with a suspected life-threatening complication were identified prospectively by a midwife or a junior doctor, who put a mark next to the women’s names in the hospital register after initial clinical examination. A research doctor (HN or two others) reviewed the hospital register three times a day, checked the medical records of newly identified suspected cases against the near-miss criteria, and determined eligibility. She used ‘data extraction forms’ to abstract a woman’s vital signs, examination results, foetal vital status, and gestational age from the records. Once an eligible woman recovered from her complication, a female interviewer (who was also a maternity staff member) obtained informed consent from her and administered a ‘woman’s questionnaire’ in the ward. It included questions on reproductive history, health care during pregnancy, plan of giving birth in a health facility, social support, and the woman’s access to money. The interviewer was encouraged to use a partition to secure privacy. In case of maternal death, the interviewer attempted to administer the questionnaire to a female relative attending the woman during her final illness before the family left the hospital. If unsuccessful, the interviewer visited the family’s residence after a mourning period, when the security risks associated with local travel were perceived as low. A male interviewer asked the woman’s husband for consent and administered a ‘husband’s questionnaire’ in the waiting area outside the ward. Questions on the timing of events were included in the husband’s questionnaire, because we expected that many women would not know the information, as some lost consciousness as a result of severe complications. Other questions addressed care-seeking from a mullah after onset of complication; careseeking costs; household assets; recent participation in community activities; the size of the social network that the husband could rely on; and village characteristics, including the presence of a vehicle, a professionally trained qabela (the Dari word for a professionally trained midwife, as opposed to daia for a traditional birth attendant), or a doctor. When respondents could not describe the timings of events with precision, they were asked to refer to specific calls to prayers (for example, the first symptom occurred after Namaz-e-Pishin but before Namaz-e-Digar), because prayers occur fairly regularly (five times a day) and play a significant role in Afghans’ lives. Questionnaires were administered in Dari, with a few exceptions conducted in Pashto. Time from admission to an in-hospital interview ranged from less than 24 h to 8 days for women and usually less than 24 h for husbands. Definitions of delays and other key variables In the maternal-health literature, the term ‘delay’ often refers to the occurrence of an excess period of time that is often assessed subjectively and coded dichotomously by an investigator. Cases may be judged ‘late’ or ‘delayed’ on the basis of the clinical conditions of the women upon reaching the hospital or cut-off points (e.g., reaching the hospital within 3 h from seizure) (Orji et al., 2002; Shaheen, Hassan, & Obaid, 2003). In our study, we define delay as the entire duration of care-seeking from the onset of complications. Therefore, all women in our study experienced a delay, which was measured without first being judged to be too long. However, we postulate that some had longer delays than others, a phenomenon that may be explained by women’s exposure to certain risk factors. Two types of delays were considered (Figs. 1 and 2). The first delay, as proposed by Thaddeus and Maine (1994), was defined as the interval between the self-reported time of first symptom

1005

recognition and the self-reported time of decision to seek care (hereafter called ‘decision delay’). A sub-category of Thaddeus and Maine’s second delay was the interval between self-reported time of decision and departure for care (hereafter called ‘departure delay’). Timings given in reference to calls to prayer were converted into 24-h-clock times by taking the midpoint between the approximate times of the two reported prayers. Small seasonal adjustments were made to reflect the changing times of prayers over the course of a year. If, by taking the midpoint, the delay was calculated to be negative (this could occur when the timing of one event was estimated by prayer times and the other event was given in exact time), then we took a conservative approach. We overestimated the delay by replacing the midpoint with the beginning or the end of the prayer time. When interviewees reported that a decision was made immediately, a delay of 5 min was assumed. Finally, delays were converted into the log scale due to skewed distribution. Remoteness was estimated by modelling the travel time from a woman’s village to the hospital using a combination of local transportation means. In Idrisi GIS software (Clark Labs, MA, USA), a grid of over 14 million cells represented the study area, with each cell representing an area of 100  100 m2. The cells were classed into five surface and three road types, with each assigned the reciprocal of the hypothetical travel speed according to the most appropriate local transportation means. The ‘cost’ function was then applied. Predicted travel times were validated against observed travel times to 12 locations, as reported by the drivers from our collaborating NGO, before we finalised the model. Admissions between December and February were coded as ‘winter’ because of heavy snowfall in these months, and admissions between June and August as ‘summer’ because June marks the beginning of the dry and hot sandstorm season. The presence of a midwife in a community was a dichotomous variable indicating the presence of a trained qabela. Indicators of access to health facilities included reported transportation cost and travel time to a first place of care. Indices of asset-based household economic status were created by adding weights equal to the inverse of the proportion of households owning selected household items and creating quartiles. The education of a woman and her husband were dichotomous variables indicating whether they had ever been to school. The woman’s relationship with her birth family was a composite measure of the answers to two questions: Whether the woman felt free to contact her birth family, and whether she had visited her relatives in the previous 3 months. ‘Strong’ indicates that the woman answered positively to both questions. ‘Moderate’ indicates she answered positively to one and negatively to the other. ‘Weak’ indicates she answered negatively to both. The woman’s relationship with her husband and his family was measured by adding scores obtained from seven dimensions of social support the woman received from members of her marital home and creating tertiles. We assessed the quality of relationships because previous studies indicated that the extent of a woman’s connectedness with her relatives might be a suitable indicator of her position in a patriarchal society (Mumtaz & Salway, 2009). Data analysis Data analysis was conducted in Stata 10.0 (Stata Corp, TX, USA). Censored normal regression techniques were used because some delays were ‘censored’ in that the true values were unknown other than being in certain intervals due to the overestimation. After bivariable analysis, we assessed whether there were interactions between complication type and other determinants presented in the conceptual framework. If interactions were plausible, we would not pool results but presented them by complication types. We applied a forward selection procedure guided by the conceptual

1006

A. Hirose et al. / Social Science & Medicine 73 (2011) 1003e1013

Distal determinants

Community characteristics Access to a vehicle Access to health personnel

Proximate determinants

Access to healthcare Transportation cost Travel time Education (woman and husband)

Practice and attitude toward healthcare Use of ANC Planned delivery in health facility Visit to a mullah

Household socioeconomic status Asset based SES

Decision delay

Remoteness Women’s status in household Relationship with birth family Relationship with husband and his family Have money at disposal

Gravidity Fig. 1. Conceptual framework to explain decision delay.

framework in our multivariable analyses (Victora, Huttly, Fuchs, & Olinto, 1997). Distal determinants were entered into the model in the same order as flows shown in the conceptual framework, and variables at the same hierarchical level were entered into the model in the order of smallest to largest p-value from likelihood ratio tests (LRT), until no further determinant would improve the model (with the threshold p-value for addition to the model being 0.1 in LRT). A determinant with p >¼0.1 in the censored regression model at the final stage was further assessed with an LRT comparing the models with and without the determinant; the determinant was dropped if the p-value in the LRT was greater than 0.1, to reach the best-fit model. Complication types were adjusted for throughout the model building process, because prior analyses indicated that they were associated with key determinants. Results were converted back to the normal scale and presented as the ratio of delay in the risk group to that in the reference group.

Ethical clearance Ethical approval was obtained from the Ministry of Public Health in Afghanistan and the London School of Hygiene & Tropical Medicine. Results Description of the achieved sample A total of 472 eligible women (including 35 women who died before discharge) were identified out of 13,927 admissions, with both female and male interviews conducted for 411 of these (87% response, Fig. 3). There were no significant demographic differences between women with a complete and incomplete set of interviews. However, difficulties in organising interviews with

Distal determinants Proximate determinants Season Access to car in the community Remoteness

Access to health facilities Transportation cost to first place of care Travel time to first place of care

Access to health l h personnel in i the community Access to a midwife Access to a doctor

Socio-economic status Asset-based socio-economic status

Departure delay

Husband’s social capital Size of the social network Participation in community activities

Husband education Fig. 2. Conceptual framework to explain departure delay.

A. Hirose et al. / Social Science & Medicine 73 (2011) 1003e1013

1007

13,927 admissions 11,701 women with no sign of major complication or gynaecological patients 2,226 examined further 1,754 1 754 women not meeting nearnear miss criteria 472 eligible women and their husbands with clinical records 21 both female and male parts of interview not conducted 219 maternal deaths 10 families left before the team made contact 1 security incident in the city 1 no consent from man (and woman’s interview withdrawn)

451 eligible women with female and/or male part of interview results

28 women’s interviews not conducted 15 maternal deaths 13 left or discharged before the team made contact

12 men’s interviews not conducted 1 maternal death (male family members were not present at a follow up visit) follow-up 8 men could not be located 3 no consent

411 p paired interviews p plus 28 male interviews and 12 female interviews Fig. 3. Study profile.

deceased women’s family members explained the differences in pregnancy outcomes, as relatives often left the hospital immediately after a death (Table 1). Table 2 shows the women’s morbidity and socio-demographic profile.

Decision delay Bivariable analyses Average durations differed significantly across complication types (Table 3), with the women with postpartum haemorrhage (PPH) and antepartum haemorrhage (APH) reporting roughly half an hour between first symptom and the decision to seek care, and others reporting durations extending to over half a day (p < 0.05). Overall, residence in remotest areas (p < 0.001), Pashtun ethnicity (p < 0.013), lack of a midwife or a vehicle in the locality (both p < 0.001), difficult access to health facilities (p ¼ 0.002 for travelling over 1.25 h and p ¼ 0.008 for costs over 500 Afghanis [equivalent to US$10]), poorest quartile (p < 0.001), an uneducated husband (p ¼ 0.014), weak relationship with birth family (p ¼ 0.029), primigravidity (p ¼ 0.009), and lack of ANC and birth plans (both p ¼ 0.001) were associated with decision delay. Weak evidence existed for the association with seeking healing from a mullah (p ¼ 0.066). A woman’s education, her relationship with

her husband’s family, and her access to money did not explain decision delay. Stratification by complication types We observed that the effect of a woman’s relationship with her birth family and, to a lesser extent, of travel time to the first place of care and the uptake of ANC was modified by complication type. The Table 1 Characteristics of women for whom both male and female interviews were completed and those for whom either one or both interviews were missing. Both interviewed One or both missing 2 p-value (N ¼ 411), % (N ¼ 61), % Demographic characteristics Women aged  20 29.4 Primigravidae 29.1 Urban population 23.8 Pregnancy profile Foetal death at admission 38.4 Stillbirthc 47.2 Maternal death 2.4 a

31.7 31.7 26.8

0.507a 0.948 0.630b

40.6 65.6 41.0

0.811 0.049 <0.001

Information on age and gravidity for one woman is missing. Information on area of residence is missing for four people from those with an incomplete interview set. c The proportion of foetuses who died in utero or during labour after the 22nd gestational week, including those never delivered due to maternal death. b

1008

A. Hirose et al. / Social Science & Medicine 73 (2011) 1003e1013

Table 2 Morbidity and socio-demographic profile of the women in the study sample (N ¼ 472). N (%) Complication types Haemorrhage during early pregnancy Antepartum haemorrhage (APH) Severe pre-eclampsia Eclampsia Impending rupture of uterus Uterine rupture Postpartum haemorrhage (PPH) Severe infection Other (heart disease and anaemia) Socio-demographic factors Mean age (years) (SD) Mean marital duration (years) (SD) (n ¼ 431) Female literacy (n ¼ 421) Reproductive factors History of stillbirth or spontaneous abortion (n ¼ 297) Median gravidity (25the75th percentile) Ethnic group (n ¼ 436) Tajik Pashtun Other

69 64 57 108 42 38 56 32 6

(14.6) (13.6) (12.1) (22.9) (8.9) (8.1) (11.9) (6.8) (1.3)

27.7 (8.0) 10.2 (6.6) 40 (9.5) 126 (42) 4 (1e7)

but evidence remained with increased delay for nonusers of ANC (AR ¼ 2.7, 95% CI ¼ 1.0e7.4). (3) Complications occurring between pregnancy week 22 and delivery The effect of lack of ANC was partly mediated through the effect of lack of birth plan but remained strong in the final model, greatly increasing the decision delay (AR ¼ 4.6, 95% CI ¼ 1.7e12.2, p ¼ 0.003). The absence of a midwife doubled the delay (AR ¼ 2.2, 95% CI ¼ 1.1e4.5, p ¼ 0.035). Seeking healing from a mullah increased the delay considerably (AR ¼ 3.2, 95% CI ¼ 1.2e8.5, p ¼ 0.018). Weak evidence suggested that a woman’s weak relationship with her birth family as well as no plan to use a health facility for the birth also increased the delay (AR ¼ 2.0, 95% CI ¼ 0.9e4.4, p ¼ 0.088, and AR ¼ 2.0, 95% CI ¼ 0.9e4.2, p ¼ 0.077, respectively). Departure delay

275 (63.1) 130 (29.8) 31 (7.1)

effects of a woman’s relationship with her birth family and of travel time to the first place of care were pronounced for women with pre-eclampsia and uterine rupture. Lack of ANC use had a clear impact on women with late pregnancy complications. Three groups of complications were therefore defined and analysed separately because the groups were not mutually exclusive: (1) Complications with clear and dramatic symptoms and that quickly develop into a fatal condition, including eclampsia, APH, and PPH. Bleeding in early pregnancy was excluded, because use of health care during pregnancy could not be assessed in many of the women with this complication, as they often reported that they had not known their pregnancy. (2) Complications with less distinctive symptoms that develop slowly, including severe pre-eclampsia, impending rupture, uterine rupture, and severe infection. (3) Complications that occurred between the 22nd gestational week and childbirth. Multivariable analyses (1) Complications with clear and dramatic symptoms Asset-based household economic status, ANC utilisation, and seeking healing from a mullah were the only variables that remained in the final model as the most important determinants of decision delay. Women in the poorest stratum experienced a delay 5.6 times greater than that of the women in the least poor stratum (95% CI ¼ 1.9e16.4, p ¼ 0.002). Non-attendance of ANC tripled the delay compared with those with a recommended minimum of four ANC visits (adjusted ratio [AR] ¼ 3.4, 95% CI ¼ 1.0e11.1, p ¼ 0.048). Seeking a mullah tripled the delay (AR ¼ 3.0, 95% CI ¼ 1.1e8.0, p ¼ 0.028) (Table 4). (2) Complications developing more slowly and with less dramatic symptoms A woman’s weak relationship with her birth family greatly increased the decision delay compared with those in a strong relationship (AR ¼ 6.4, 95% CI ¼ 2.7e15.3, p < 0.001). Most of the effect of ANC was mediated through the plan to use a health facility,

Bivariable analyses Compared to the decision delay, the differences in departure delay across complication types were small (Table 5). The delay was approximately 1 h for all complication types except for severe infection (p < 0.001). Remoteness, access to a vehicle, access to health personnel and facilities, household socioeconomic status, social network, and community participation were each associated with departure delay in the expected direction. Admissions in winter and summer were associated with increased delay (p ¼ 0.048 and 0.070, respectively). Multivariable analyses Since there was no evidence of interactions, results were pooled. In the final model, access to first place of care was found to be a very important determinant of the departure delay. Women who travelled for more than 1 h 15 min to a first point of care had a 2.2 times longer delay than those reaching a first point of care within half an hour (95% CI ¼ 1.5e3.2, p < 0.001). Women whose initial transportation costs were greater than 500 Afghanis had a 1.8 times longer delay than those whose initial cost was nil (95% CI ¼ 1.1e2.7, p ¼ 0.013). The size of the husband’s social network was also important: Those who had no one to rely on in time of long-term crisis took 1.8 times longer to leave home after decision (95% CI ¼ 1.1e3.1, p ¼ 0.032), compared with respondents who could rely on five or more people. Lack of access to a doctor in the community also significantly increased the delay, by a ratio of 1.7 (95% CI ¼ 1.2e2.4, p ¼ 0.001). Finally, women admitted in summer and in winter experienced a longer delay than those admitted in spring (AR ¼ 1.6, 95% CI ¼ 1.0e2.4, p ¼ 0.030, and AR ¼ 1.5, 95% CI ¼ 1.0e2.2. p ¼ 0.060, respectively) (Table 6). Discussion The study shows considerable differences in explanatory models for the decision and departure delays. While the decision delay was mainly explained by exposure to health care services, household economic status, and a woman’s household status, the departure delays were largely explained by geographical accessibility to health care, seasons, and the size of the husband’s social network. In addition, the duration and determinants of a decision delay were strongly influenced by the nature of the symptoms. When symptoms were clear, poverty was associated with longer decision delay. When symptoms were less obvious, weak relationships with birth family were associated with delayed decision. For complications occurring after the 22nd gestation week before

A. Hirose et al. / Social Science & Medicine 73 (2011) 1003e1013

1009

Table 3 Association between risk factors and the delay between symptom recognition and decision to seek care.

Complication type

Remoteness

Ethnic group

Midwife in community Vehicle in community Travel time to 1st place

Travel cost to 1st place

Asset-based household economic status

Husband’s education Woman’s education Woman’s relationship with birth family Woman’s relationship with husband and his family Have money at her disposal Gravidity

ANC

Birth plan to use a health facility

Visited a mullah after onset of complication

N (%)

Delay in hours

Ratio to the reference group [95% CI]

PPH (ref) APH Eclampsia Bleeding in early pregnancy Severe pre-eclampsia Rupture of uterus Impending rupture Severe infection Other Least remote (ref) 3rd most remote 2nd most remote Most remote Tajik (ref) Pashtun Other Yes (ref) No Yes (ref) No 0.5 h (ref) 1.25 & >0.5 h >1.25 h 0 Afs (ref) 100 & >0 Afs 500 & >100 Afs >500 Afs Least poor (ref) 3rd poorest 2nd poorest Poorest Yes (ref) No Yes (ref) No Strong (ref) Moderate Weak Strong (ref) Moderate Weak Yes (ref) No 6 or more (ref) 2e5 pregnancies Primigravidae 4 or more (ref) 1e3 times 0 Yes (ref) No

48 (11.4) 61(14.5) 98 (23.2) 60 (14.2) 50 (11.8) 32 (7.6) 39 (9.2) 28 (6.6) 5 (1.2) 116 (27.7) 91(21.7) 112 (26.7) 100 (23.9) 264 (63.2) 124 (29.7) 30(7.2) 128 (32.1) 271(67.9) 367 (88.2) 49 (11.8) 186 (46.9) 80 (20.2) 131 (33.0) 118 (29.4) 88 (21.9) 101(25.1) 95 (23.6) 103 (24.5) 106 (25.2) 87 (20.7) 124 (29.5) 147 (35.2) 271 (64.8) 51 (12.9) 343 (87.1) 130 (34.4) 147 (38.9) 101 (26.7) 105 (27.7) 131 (34.5) 143 (37.7) 105 (27.7) 274 (72.3) 162 (38.5) 132 (31.4) 127 (30.2) 86 (25.5) 154 (45.7) 97 (28.8) 95 (28.3) 241 (71.7)

0.5 0.6 1.2 1.3 2.4 2.9 4.5 16.6 15.9 0.7 1.1 1.7 4 1.2 2.5 1.8 0.7 2 1.3 4.8 1 1.6 2.7 1.1 0.7 1.8 3.1 0.5 1.3 3.3 2.7 1 1.9 1.3 1.5 1 1.3 2.2 1.2 1.1 1.8 1.1 1.5 0.9 1.9 2.2 0.8 1.4 2.9 0.6 2.1

1.0 1.2 [0.5e3.3] 2.5 [1.0e6.2] 2.8 [1.0e7.4] 5.3 [1.9e14.8] 6.2 [2.0e19.7] 9.7 [3.3e28.7] 35.6 [10.9e116.4] 34.4 [3.4e348.7] 1.0 1.5 [0.7e3.2] 2.3 [1.1e4.6] 5.4 [2.6e11.3] 1.0 2.1 [1.2e3.8] 1.5 [0.5e4.4] 1.0 2.9 [1.6e5.3] 1.0 3.8 [1.7e8.6] 1.0 1.6 [0.8e3.2] 2.6 [1.4e4.8] 1.0 0.6 [0.3e1.3] 1.6 [0.8e3.4] 2.8 [1.3e5.9] 1.0 2.6 [1.3e5.5] 6.9 [3.2e14.9] 5.6 [2.7e11.4] 1.0 2.0 [1.2e3.5] 1.0 1.2 [0.5e2.6] 1.0 1.3 [0.7e2.5] 2.3 [1.1e4.7] 1.0 0.9 [0.4e1.8] 1.5 [0.7e3.0] 1.0 1.4 [0.7e2.6] 1.0 2.0 [1.1e3.9] 2.4 [1.2e4.5] 1.0 1.9 [0.9e4.0] 3.9 [1.8e8.7] 1.0 3.2 [1.7e6.1]

No (ref) Yes

340 (83.7) 66 (16.3)

1.4 2.7

1.0 2.0 [1.0e4.1]

p-value from censored regression 0.688 0.047 0.042 0.002 0.002 <0.001 <0.001 0.003 0.291 0.024 <0.001 0.013 0.416 <0.001 0.001 0.213 0.002 0.214 0.211 0.008 0.010 <0.001 <0.001 0.014 0.729 0.464 0.029 0.758 0.262 0.319 0.028 0.009 0.079 0.001 0.001

0.066

Note: (ref) ¼ Reference group.

childbirth, the absence of a midwife in the community increased the decision delay. ANC uptake was associated with shorter decision delay in all the complication groups. Possible explanations of key findings Decision delay When their mobility is restricted, women may need to negotiate with unwilling and uncooperative family members to make the journey to obtain health care. Negotiating for health care for illness with minor symptoms may be particularly difficult, and a woman may need to manipulate her available resources to justify travel and costs. A strong relationship with her birth family

may empower her, as birth family members may be able to accompany her to a health facility, take care of children, do other domestic tasks, or bear health care costs. A close relationship with her birth family may paradoxically reflect a woman’s strong position in her marital home (Dyson & Moore, 1983; Mumtaz, 2002; Unnithan-Kumar, 1999). We found contrasting results with regards to trained midwives in communities and seeking care from a mullah: Having such a midwife in the locality reduced decision-making time, while seeking healing from a mullah prolonged it. Although we did not elucidate how access to a midwife influences decision-making, other studies have shown the important role of professionally trained village midwives in detecting obstetric complications and

1010

A. Hirose et al. / Social Science & Medicine 73 (2011) 1003e1013

Table 4 Results of multivariable censored normal regression models to explain the duration of decision-making time, by complication group. Group 1: Clear symptoms AR [95% CI] Remoteness

Midwife in community Asset-based household economic status

Woman's relationship with birth family

Gravidity

ANC

Visited a mullah Birth plan to use a facility

Least (ref) 3rd most remote 2nd most remote Most remote Yes (ref) No Least poor (ref) 3rd poorest 2nd poorest Poorest Strong (ref) Moderate Weak 6 or more (ref) 2e5 pregnancies Primigravidae 4 or more visits (ref) 1e3 visits 0 Did not visit (ref) Visited a mullah Planned (ref) Did not plan

p-value

Group 3: After week 22

AR [95% CI]

AR [95% CI]

1 2.8 [1.07.7] 0.9 [0.32.6] 2.0 [0.76.2]

Dropped Dropped 1 3.0 [1.18.7] 3.7 [1.112.2] 5.6 [1.916.4]

0.039 0.032 0.002

N/A

N/A 1 2.5 [0.97.5] 3.4 [1.011.1] 1 3.0 [1.18.0] N/A

Group 2: Less dramatic symptoms

0.091 0.048 0.028

N/A 1 1.3 [0.53.6] 2.4 [0.87.3] 0.7 [0.22.4] 1 2.5 [1.15.5] 6.4 [2.715.3] 1 1.3 [0.62.9] 2.2 [0.95.2] 1 1.2 [0.52.5] 2.7 [1.07.4] N/A 1 2.7 [1.25.9]

p-value 0.048 0.908 0.226

0.625 0.126 0.528 0.025 <0.001 0.452 0.075 0.727 0.056

0.013

Dropped 1 2.2 [1.14.5] 1 1.8 [0.74.4] 2.7 [0.97.7] 1.6 [0.64.6] 1 1.8 [0.84.0] 2.0 [0.94.4]

N/A 1 2.6 [1.15.8] 4.6 [1.712.2] 1 3.2 [1.28.5] 1 2.0 [0.94.2]

p-value

0.035 0.219 0.063 0.382 0.138 0.088

0.022 0.003 0.018 0.077

Note: (ref) ¼ Reference group. These are results of three independent models. N/A indicates that the variable was not entered into the model because it had no association with delay in bivariable analysis and/or the result of LRT was >0.1. ‘Dropped’ indicates that the variable was entered at an earlier stage of the model building but was dropped from the final model because its effect was mediated through other variables.

sensitising communities to the urgency of obstetric emergencies (D’Ambruoso et al., 2009). The closure of midwifery schools during the Taliban regime resulted in a severe lack of SBAs, but a large cadre of midwives has since been trained and deployed to facilities (Currie, Azfar, & Fowler, 2007). Our findings underline the importance of midwives and other health workers working closely with communities to increase health care utilisation and potentially reduce maternal mortality. Our results suggest that ANC consultations may lead to prompt decision-making in an obstetric emergency. Although the study cannot explain the mechanism of behaviour change, women may gain knowledge about the risks associated with delivery and start making plans for childbirth. Communications with a midwife during ANC consultations may help build trust in health care providers. Our findings underline the important role of ANC as a linkage to delivery care, including EmOC. However, caution is needed in interpreting this finding: Women seeking EmOC promptly may already be more risk-aware and more likely to utilise ANC. Departure delay Our finding that distance from facilities increases the departure delay is worrisome, given that obstetric complications can rapidly lead to death. There are several possible explanations. Individuals unfamiliar with health facilities might ask neighbours, friends, and relatives for advice on where to go for care. In addition, the further away one lives from facilities, the more financial, logistical, and emotional preparations might be necessary. Previously, the time to obtain transportation was quantified to describe care-seeking behaviours in rural Matlab, Bangladesh (Killewo, Anwar, Bashir, Yunus, & Chakraborty, 2006), but associations between the time to obtain transportation and distance were not specifically addressed. Social networks also play an important role in facilitating access to health care. Many rural residents do not have a regular income, and they borrow cash from kin and peers for health care (authors’

unpublished data; Steinhardt et al., 2009). Members of a social network may provide information about where to obtain health care and encourage good care-seeking behaviours. Women reported a longer departure delay in the summer and winter than in the spring. Variations in the magnitude of maternal morbidity and mortality due to problems associated with care-seeking in specific seasons have been reported in other countries (Anya, 2004; Etard, Kodio, & Ronsmans, 2003; Hounton et al., 2008; Lema, Changole, Kanyighe, & Malunga, 2005). Agricultural work and difficulty in travelling explained the increased number of institutional maternal deaths during the rainy planting seasons in Malawi (Lema et al., 2005). The easier access and low opportunity costs to seek care explained the increased institutional maternal deaths and complications during the dry season in Burkina Faso (Hounton et al., 2008). While the delays during the Afghan winter may be explained by difficulties in travelling in snow, the summer delays can be explained by a broader range of factors. The speed of life slows down because of the intense heat, and it may take more time to mobilise resources and arrange travel. Given the summer’s farming and other labour requirements, women may find it difficult to seek care outside their village in the absence of a male family member. The increased level of offensives against foreign forces and allies in rural Afghanistan during the summer may elevate poor residents’ fear of opportunistic robberies and lead to delay their journeys. Strengths and limitations We focused on ‘near-miss upon arrival’ cases instead of maternal deaths alone. The sample allowed for stratification by complication types and the comparison of long delays to short delays, while controlling for confounders. We identified the determinants of care-seeking delays for different complication groups. Separating a care-seeking duration into two distinct time periods helped to explain the mechanisms of particular delays.

A. Hirose et al. / Social Science & Medicine 73 (2011) 1003e1013

1011

Table 5 Association between risk factors and the delay between decision to seek care and departure from home.

Complication type

Remoteness

Ethnic group

Season

Vehicle in community Doctor in community Midwife in community Asset-based household economic status

Husband's education Travel cost to 1st place of care

Travel time to 1st place of care

No. people husband can borrow a small amount of money from

No. people husband can rely on in long-term emergency

Participated in community activities in last 12 months

N (%)

Delay in hours

Ratio to the reference group [95% CI]

APH (ref) Severe pre-eclampsia Eclampsia PPH Uterine rupture Bleeding in early pregnancy Impending rupture of uterus Severe infection Other Least remote (ref) 3rd most remote 2nd most remote Most remote Tajik (ref) Pashtun Other Spring (ref) Summer Autumn Winter Yes (ref) No Yes (ref) No Yes (ref) No Least poor (ref)

60 (14.3) 47 (11.2) 99 (23.6) 48 (11.4) 34 (8.1) 59 (14.0) 40 (9.5) 28 (6.7) 5 (1.2) 116 (27.8) 90 (21.6) 114 (27.3) 97 (23.3) 260 (62.5) 125 (30.0) 31 (7.5) 113 (27.0) 88 (21.0) 111 (26.5) 107 (25.5) 365 (88.2) 49 (11.8) 189 (46.1) 221 (53.9) 126 (31.8) 270 (68.2) 102 (24.4)

0.8 0.8 0.9 0.9 1.2 1.3 1.7 4.1 5.6 0.7 1.3 0.8 2.6 1.1 1.2 1.5 0.9 1.3 1.1 1.3 1.0 3.0 0.7 1.7 0.8 1.4 0.6

1.0 1.0 [0.51.9] 1.1 [0.71.9] 1.2 [0.62.2] 1.6 [0.83.1] 1.7 [1.03.1] 2.3 [1.24.3] 5.2 [2.511.1] 7.2 [1.631.9] 1.0 1.9 [1.22.9] 1.2 [0.81.9] 3.8 [2.55.9] 1.0 1.1 [0.81.6] 1.4 [0.82.6] 1.0 1.5 [1.02.5] 1.3 [0.82.0] 1.6 [1.02.5] 1.0 3.0 [1.95.0] 1.0 2.5 [1.83.5] 1.0 1.9 [1.32.7] 1.0

3rd poorest 2nd poorest Poorest Yes (ref) No 0 Afs (ref) <¼100 & > 0 Afs <¼ 500 & > 100 Afs > 500 Afs <¼ 0.5 hour (ref) <¼ 1.25 & > 0.5 hr > 1.25 hr 5 or more (ref)

106 (25.4) 86 (20.6) 124 (29.7) 147 (35.3) 269 (64.7) 119 (29.8) 85 (21.3) 100 (25.0) 96 (24.0) 189 (47.6) 82 (20.7) 126 (31.7) 100 (24.4)

1.0 1.1 2.1 0.8 1.4 0.7 0.8 1.2 2.4 0.7 0.9 2.5 0.9

1.8 [1.12.7] 1.8 [1.22.9] 3.6 [2.35.5] 1.0 1.7 [1.22.4] 1.0 1.1 [0.71.7] 1.7 [1.12.6] 3.4 [2.25.3] 1.0 1.3 [0.91.9] 3.5 [2.54.9] 1.0

0.012 0.010 <0.001

3e4 1e2 0 5 or more (ref)

145 145 20 35

(35.4) (35.4) (4.9) (8.5)

1.1 1.4 1.2 0.7

1.3 [0.92.0] 1.6 [1.12.5] 1.4 [0.63.2] 1.0

0.212 0.027 0.401

3e4 1e2 0 Yes (ref)

21 142 214 134

(5.1) (34.5) (51.9) (32.8)

0.7 1.0 1.4 0.8

0.9 [0.42.3] 1.3 [0.72.5] 1.9 [1.13.6] 1.0

0.860 0.357 0.032

No

275 (67.2)

1.4

1.7 [1.22.4]

0.002

The study had limitations, however. We could not include women with life-threatening complications who did not reach the study hospital. Some may have died at home without seeking care, or at lower-level facilities, or on the way. Others may have survived despite not seeking adequate care. Since women who never made the decision and women who died possibly had longer delays than study participants, the magnitude of certain determinants may have been underestimated (although the contrary is possible). Factors determining positive deviant behaviours of women seeking care before developing a near-miss event could not be identified in this study. On the other hand, if women survived by luck without any medical care, their survival was, by definition, a randomly occurring unpredictable phenomenon. The exclusion of the ‘randomly chosen’ would not have resulted in biased estimates of the associations.

p-value from censored regression 0.952 0.612 0.565 0.199 0.064 0.014 <0.001 0.010 0.006 0.317 <0.001 0.519 0.290 0.070 0.314 0.048 <0.001 <0.001 0.001

0.002 0.755 0.018 <0.001 0.191 <0.001

The number of near-miss events among unmarried women would have been so small in this population that their exclusion would not have altered the observed associations. A separate qualitative study focussing on this sub-group may be necessary to highlight specific barriers. There may be measurement errors in estimating delays. Timings were assessed retrospectively, mostly subjectively. There may have been delays in symptom recognition, which were not assessed in our study, particularly for complications with subtle symptoms. A prospective study of pregnant women to quantify delays among those developing complications would reduce the measurement errors, but such a study would be unethical without ensuring a safety procedure. Departure delay and decision delay were assessed separately for their determinants. Nonetheless, dividing a continuous

1012

A. Hirose et al. / Social Science & Medicine 73 (2011) 1003e1013

Table 6 Results of multivariable censored regression models to explain the duration of time from decision to seek care to departure from home. Delay Ratio to the p-value from reference group censored [95% CI] regression model Season

Spring (ref) Summer Autumn Winter Doctor in community Yes (ref) No Travel time to 1st place <¼ 0.5 hour (ref) of care <¼ 1.25 & > 0.5 hr >1.25 hr Travel cost to 1st place Didn't pay (ref) of care <¼100 & > 0 Afs <¼ 500 & > 100 Afs >500 Afs No. people husband 5 or more (ref) can rely on in longterm emergency 3e4 1e2 0

1 1.6 1.2 1.5 1 1.7 1

[1.02.4] [0.81.8] [1.02.2]

0.030 0.281 0.060

[1.22.4]

0.001

0.9 [0.61.4] 2.2 [1.53.2] 1

0.754 <0.001

1.0 [0.61.6] 1.3 [0.81.9]

0.997 0.280

1.8 [1.12.7] 1

0.013

1.0 [0.52.2] 1.4 [0.82.5] 1.8 [1.13.1]

0.994 0.220 0.032

Note: (ref) ¼ Reference group. Complication types and remoteness have been adjusted for.

process into two durations may appear somewhat artificial, as the latter delay may significantly factor into the former. Our results may simply reflect the complex process of making a decision: After the need for health care has been recognised and the decision to seek care made, logistical barriers must be overcome before a woman starts the journey to a health care provider. We built independent models to identify determinants of decision delay for three groups of women. Those wishing to identify determinants for a specific disease may find it difficult to do so from our results because the grouping was not mutually exclusive. Our focus was to explore how determinants varied by the clarity and timing of symptoms and to present a woman’s perspective. Conclusions How best to deliver obstetric care to reduce maternal mortality remains a contentious issue. Our findings are relevant in regions where universal coverage of facility-based intrapartum care is unlikely in the near future, raising again questions regarding the importance of ANC. Birth-preparedness and complicationreadiness strategies have been adopted in many countries with high MMR, as part of a focused ANC strategy with community- or clinic-based behaviour change communication activities, social mobilisation, or advocacy as components (WHO, 2006). However, most studies evaluating these interventions with maternal-health end points have not been robust in design (Stanton, 2004), and evidence of their impact on decision-making for EmOC hardly exists. A robust evaluation of the effectiveness of a clinic-based, birth-preparedness intervention during ANC on seeking EmOC may be worthwhile. The study also highlights that an Afghan woman’s poor social position, marked by separation from her natal family, could adversely affect utilisation of obstetric care. The importance of family support for women during pregnancy and childbirth could be stressed further in safe motherhood programmes. Community-

based interventions to increase social support for women (for example, by empowering community-based women’s groups (Manandhar et al., 2004)) may prove effective in Afghanistan. As Afghan women’s status is likely to remain poor for the foreseeable future, targeting senior male and female household members and key stakeholders (including mullahs) in behaviour-change communication activities may be crucial to social change. Our findings also suggest that strengthening men’s social networks may help facilitate access to health care. Finally, in settings where the coverage of SBA is low, a reduction of maternal mortality may be achieved if women with complications accessed and obtained EmOC immediately. From a policy perspective, it would be preferable if those living far from facilities left home quickly, yet the study showed the contrary was true. Introducing a financial mechanism (such as a voucher or conditional cash transfer) to provide EmOC services and transportation to EmOC facilities free of charge or at a highly subsidised rate may help (Lahariya, 2009). More generally, upgrading and expanding roads and facilities are imperative. The global health community should continue to commit to ensuring equitable access to health facilities in Afghanistan. Acknowledgements This project was funded by World Vision Japan. We would like to thank Dr. Saida Said and the doctors and midwives of the maternity ward of Herat Regional Hospital, who collaborated with us during fieldwork. AH was supported by the Joint Japan/World Bank Graduate Scholarship Program. References Adisasmita, A., Deviany, P. E., Nandiaty, F., Stanton, C., & Ronsmans, C. (2008). Obstetric near miss and deaths in public and private hospitals in Indonesia. BMC Pregnancy Childbirth, 8(10), 10. Amowitz, L. L., Reis, C., & Iacopino, V. (2002). Maternal mortality in Herat Province, Afghanistan, in 2002: an indicator of women’s human rights. JAMA: the Journal of the American Medical Association, 288(10), 1284e1291. Anya, S. E. (2004). Seasonal variation in the risk and causes of maternal death in the Gambia: Malaria appears to be an important factor. American Journal of Tropical Medicine and Hygiene, 70(5), 510e513. Barnes-Josiah, D., Myntti, C., & Augustin, A. (1998). The ’three delays’ as a framework for examining maternal mortality in Haiti. Social Science & Medicine, 46(8), 981e993. Bartlett, L. A., Mawji, S., Whitehead, S., Crouse, C., Dalil, S., Ionete, D., et al. (2005). Where giving birth is a forecast of death: maternal mortality in four districts of Afghanistan, 1999-2002. Lancet, 365(9462), 864e870. Campbell, O. M., & Graham, W. J. (2006). Strategies for reducing maternal mortality: getting on with what works. Lancet, 368(9543), 1284e1299. Chowdhury, M. E., Ahmed, A., Kalim, N., & Koblinsky, M. (2009). Causes of maternal mortality decline in Matlab, Bangladesh. Journal of Health, Population, and Nutrition, 27(2), 108e123. Currie, S., Azfar, P., & Fowler, R. C. (2007). A bold new beginning for midwifery in Afghanistan. Midwifery, 23(3), 226e234. D’Ambruoso, L., Achadi, E., Adisasmita, A., Izati, Y., Makowiecka, K., & Hussein, J. (2009). Assessing quality of care provided by Indonesian village midwives with a confidential enquiry. Midwifery, 25(5), 528e539. Dyson, T., & Moore, M. (1983). On kinship structure, female autonomy, and demographic behavior in India. Population and Development Review, 9(1), 35e60. Etard, J. F., Kodio, B., & Ronsmans, C. (2003). Seasonal variation in direct obstetric mortality in rural Senegal: Role of malaria? American Journal of Tropical Medicine and Hygiene, 68(4), 503e504. Filippi, V., Ronsmans, C., Gohou, V., Goufodji, S., Lardi, M., Sahel, A., et al. (2005). Maternity wards or emergency obstetric rooms? Incidence of near-miss events in African hospitals. Acta Obstetricia et Gynecologica Scandinavica, 84(1), 11e16. Filippi, V., Richard, F., Lange, I., & Ouattara, F. (2009). Identifying barriers from home to the appropriate hospital through near-miss audits in developing countries. Best Practice & Research. Clinical Obstetrics & Gynaecology, 23(3), 389e400. Fournier, P., Dumont, A., Tourigny, C., Dunkley, G., & Drame, S. (2009). Improved access to comprehensive emergency obstetric care and its effect on institutional maternal mortality in rural Mali. Bulletin of the World Health Organization, 87(1), 30e38. Hansen, P. M., Peters, D. H., Niayesh, H., Singh, L. P., Dwivedi, V., & Burnham, G. (2008). Measuring and managing progress in the establishment of basic health

A. Hirose et al. / Social Science & Medicine 73 (2011) 1003e1013 services: the Afghanistan health sector balanced scorecard. International Journal of Health Planning and Management, 23(2), 107e117. Hounton, S. H., Sombie, I., Townend, J., Ouedraogo, T., Meda, N., & Graham, W. J. (2008). The tip of the iceberg: evidence of seasonality in institutional maternal mortality and implications for health resources management in Burkina Faso. Scandinavian Journal of Public Health, 36(3), 310e317. Killewo, J., Anwar, I., Bashir, I., Yunus, M., & Chakraborty, J. (2006). Perceived delay in healthcare-seeking for episodes of serious illness and its implications for safe motherhood interventions in rural Bangladesh. Journal of Health, Population and Nutrition, 24(4), 403e412. Koblinsky, M., Matthews, Z., Hussein, J., Mavalankar, D., Mridha, M. K., Anwar, I., et al. (2006). Going to scale with professional skilled care. Lancet, 368(9544),1377e1386. Lahariya, C. (2009). Cash incentives for institutional delivery: linking with antenatal and post natal care may ensure ’continuum of care’ in India. Indian Journal of Community Medicine, 34(1), 15e18. Lema, V. M., Changole, J., Kanyighe, C., & Malunga, E. V. (2005). Maternal mortality at the Queen Elizabeth Central Teaching Hospital, Blantyre, Malawi. East African Medical Journal, 82(1), 3e9. Manandhar, D. S., Osrin, D., Shrestha, B. P., Mesko, N., Morrison, J., Tumbahangphe, K. M., et al. (2004). Effect of a participatory intervention with women’s groups on birth outcomes in Nepal: cluster-randomised controlled trial. Lancet, 364(9438), 970e979. Mayhew, M., Hansen, P. M., Peters, D. H., Edward, A., Singh, L. P., Dwivedi, V., et al. (2008). Determinants of skilled birth attendant utilization in Afghanistan: a cross-sectional study. American Journal of Public Health, 98(10), 1849e1856. MOH. (2003). A basic package of health services for Afghanistan. Kabul: Ministry of Health, Transitional Islamic Government of Afghanistan. Mumtaz, Z. (2002) Gender and reproductive health. A need for conceptualisation (PhD thesis). London: University of London. Mumtaz, Z., & Salway, S. (2009). Understanding gendered influences on women’s reproductive health in Pakistan: moving beyond the autonomy paradigm. Social Science & Medicine, 68(7), 1349e1356. Okonofua, F. E., Abejide, A., & Makanjuola, R. A. (1992). Maternal mortality in Ile-Ife, Nigeria: A study of risk factors. Studies in Family Planning, 23(5), 319e324. Okusanya, B. O., Okogbo, F. O., Momoh, M. M., Okogbenin, S. A., & Abebe, J. O. (2007). Maternal mortality and delay: Socio-demographic characteristics of maternal deaths with delay in Irrua, Nigeria. Nigerian Journal of Medicine, 16(1), 38e41. Orji, E. O., Ogunlola, I. O., & Onwudiegwu, U. (2002). Brought-in maternal deaths in south-west Nigeria. Journal of Obstetrics and Gynaecology, 22(4), 385e388.

1013

Paxton, A., Maine, D., Freedman, L., Fry, D., & Lobis, S. (2005). The evidence for emergency obstetric care. International Journal of Gynaecology and Obstetrics, 88(2), 181e193. Prual, A., Bouvier-Colle, M. H., de Bernis, L., & Breart, G. (2000). Severe maternal morbidity from direct obstetric causes in West Africa: incidence and case fatality rates. Bulletin of the World Health Organization, 78(5), 593e602. Roost, M., Altamirano, V. C., Liljestrand, J., & Essen, B. (2009). Priorities in emergency obstetric care in BoliviadMaternal mortality and near-miss morbidity in metropolitan La Paz. BJOG: an International Journal of Obstetrics and Gynaecology, 116(9), 1210e1217. Say, L., Souza, J. P., & Pattinson, R. C. (2009). Maternal near missdTowards a standard tool for monitoring quality of maternal health care. Best Practice & Research. Clinical Obstetrics & Gynaecology, 23(3), 287e296. Shaheen, B., Hassan, L., & Obaid, M. (2003). Eclampsia, a major cause of maternal and perinatal mortality: a prospective analysis at a tertiary care hospital of Peshawar. Journal of the Pakistan Medical Association, 53(8), 346e350. Smith, D. J. (2009a). Community-based dispute resolution process in Nangarhar province. Kabul: AREU. Smith, D. J. (2009b). Decisions, desires and diversity: marriage practices in Afghanistan. Kabul: AREU. Stanton, C. K. (2004). Methodological issues in the measurement of birth preparedness in support of safe motherhood. Evalucation Review, 28(3), 179e200. Steinhardt, L. C., Waters, H., Rao, K. D., Naeem, A. J., Hansen, P., & Peters, D. H. (2009). The effect of wealth status on care seeking and health expenditures in Afghanistan. Health Policy and Planning, 24(1), 1e17. Supratikto, G., Wirth, M. E., Achadi, E., Cohen, S., & Ronsmans, C. (2002). A districtbased audit of the causes and circumstances of maternal deaths in South Kalimantan, Indonesia. Bulletin of the World Health Organization, 80(3), 228e234. Thaddeus, S., & Maine, D. (1994). Too far to walk: maternal mortality in context. Social Science & Medicine, 38(8), 1091e1110. Unnithan-Kumar, M. (1999). Households, kinship and access to reproductive health care among rural Muslim women in Jaipur. Economic and Political Weekly, 34(10-11), 621e630. Urassa, E., Massawe, S., Lindmark, G., & Nystrom, L. (1997). Operational factors affecting maternal mortality in Tanzania. Health Policy and Planning, 12(1), 50e57. Victora, C. G., Huttly, S. R., Fuchs, S. C., & Olinto, M. T. A. (1997). The role of conceptual frameworks in epidemiological analysis: A hierarchical approach. International Journal of Epidemiology, 26(1), 224e227. WHO. (2006). Standards for maternal and neonatal care. Geneva: WHO.