The evolving contribution of non-communicable diseases to maternal mortality in Jamaica, 1998-2015: a population-based study.

OBJECTIVE: To describe trends in indirect cause-specific pregnancy-related mortality from 1998 to 2015. DESIGN: Secondary analysis of annual, national cross-sectional database of maternal and late maternal deaths, identified through active surveillance of deaths among women aged 10-50 years. SETTING: Jamaica, a middle-income Caribbean country. POPULATION: Maternal and late maternal deaths. METHODS: Descriptive trend analyses of demographic and cause-specific maternal and pregnancy-related mortality ratios undertaken comparing the periods 1998-2003, 2004-2009 and 2010-2015. Multivariate logistic regression was used to confirm changes in risk of indirect death. MAIN OUTCOME MEASURES: Maternal, pregnancy-related, direct, indirect and cause-specific mortality ratios (deaths/100 000 live births). RESULTS: Maternal deaths from indirect conditions increased between the first two periods (P = 0.004) and stabilised in the third (P = 0.085). Associated with upward movement in cardiovascular deaths (P[trend] = 0.003), women under 25 years were at elevated risk (odds ratio 1.44, 95% CI 1.00-2.08; P = 0.052). Haematological/immunological conditions (69% sickle cell disease) ranked second but did not vary with time. Health service utilisation was similar across age, parity, health region and major cause categories (non-communicable diseases, non-obstetric infections, direct), however women with indirect conditions spent more time in hospital (median 5 days versus 3 days) and more often died after the puerperium. CONCLUSIONS: Medical conditions, especially cardiovascular disease, are increasingly associated with maternal and late maternal mortality. Middle-income countries need to simultaneously improve management of indirect conditions, while redoubling efforts to reduce direct deaths. Postpuerperal medical services should be integrated into routine infant health services to improve continuity of care during this high-risk period. TWEETABLE ABSTRACT: Maternal survival (SDG 3.1) in LMICs requires better care for women with both non-communicable diseases and obstetric conditions.

The Quality and Completeness of 2008 Perinatal and Under-five Mortality Data from Vital Registration, Jamaica.

OBJECTIVE: To evaluate the completeness and timeliness of registration of stillbirths and under-five deaths and the validity of the certification and coding process. SUBJECTS AND METHODS: Registered stillbirths and under-five deaths occurring in 2008 were compared to hospital, police, forensic pathologist and coroner's records. Missed cases and new information such as birthweight, gestation and date of birth were added to the database. A 10% random sample was evaluated to measure the quality of certification and coding. RESULTS: Of 646 stillbirths [≥ 1000 g] and 933 under-five deaths, 69% and 79%, respectively were registered by December 31, 2009, for inclusion in the 2008 final demographic returns. Non-reporting of stillbirths was associated with infant gender, region and place of death (seven of 21 public hospitals accounted for 96% of unregistered stillbirths). Among under-five deaths, age at death, region, place and cause of death were important. Injury and community deaths increased with age. Registration delays including non-registration were associated with coroner's inquests. Most [80%] stillbirth certificates lacked usable cause of death data. Neonatal deaths due to prematurity and perinatal asphyxia were often misclassified by coders. The stillbirth [≥ 1000 g], infant and under-five mortality rates were 15, 20 and 22/1000 births/live births, respectively. CONCLUSIONS: While registration of stillbirths and under-five deaths improved between 1998 and 2008, persistent under-reporting reduced official rates by 20-31%. A new perinatal death certificate documenting maternal and fetal causes of death and risk factors such as birthweight, gestation and age at death would improve stillbirth and neonatal death (0-28 days) data quality.

Impact of the Jamaican birth cohort study on maternal, child and adolescent health policy and practice.

The Jamaica Perinatal Morbidity and Mortality Survey (JPMMS) was a national study designed to identify modifiable risk factors associated with poor maternal and perinatal outcome. Needing to better understand factors that promote or retard child development, behaviour and academic achievement, we conducted follow-up studies of the birth cohort. The paper describes the policy developments from the JPMMS and two follow-up rounds. The initial study (1986-87) documented 94% of all births and their outcomes on the island over 2 months (n = 10 508), and perinatal (n = 2175) and maternal deaths (n = 62) for a further 10 months. A subset of the birth cohort, identified by their date of birth through school records, was seen at ages 11-12 (n = 1715) and 15-16 years (n = 1563). Findings from the initial survey led to, inter alia, clinic-based screening for syphilis, referral high-risk clinics run by visiting obstetricians, and the redesign and construction of new labour wards at referral hospitals. The follow-up studies documented inadequate academic achievement among boys and children attending public schools, and associations between under- and over-nutrition, excessive television viewing (>20 h/week), inadequate parental supervision and behavioural problems. These contributed to the development of a television programming code for children, a National Parenting Policy, policies aimed at improving inter-sectoral services to children from birth to 5 years (Early Childhood Commission) and behavioural interventions of the Violence Prevention Alliance (an inter-sectoral NGO) and the Healthy Lifestyles project (Ministry of Health). Indigenous maternal and child health research provided a local evidence base that informed public policy. Collaboration, good communication, being vigilant to opportunities to influence policy, and patience has contributed to our success.

Perinatal factors in students admitted to the University of the West Indies data from the Jamaican Perinatal Study.

OBJECTIVE: To compare perinatal and social factors in students admitted to The University of the West Indies (UWI), Kingston, Jamaica, at age 18 years with those in the rest of the Jamaican Perinatal Cohort. METHOD: The Jamaican Perinatal Survey recorded demographic and perinatal details in 10 527 or 97% of births in Jamaica in September and October 1986. Eighteen years later 140 of these were admitted to the UWI in Kingston, Jamaica. The perinatal features of these UWI students have been compared with the rest of the Perinatal Survey Cohort. RESULTS: Mothers of UWI students were older (p < 0.001), more likely to be married at the time of birth (p < 0.001), had earlier and more complete antenatal care (p < 0.05) and greater educational achievement at time of pregnancy (p < 0.001). These mothers of UWI students were also more likely to have diabetes (p < 0.01), operative deliveries (p < 0.01) and to attend private hospitals (p < 0.01). The UWI students had fewer siblings by their mothers (p < 0.05), were less likely to be low birthweight babies (p = 0.035) and more likely to be full-term (37-42 weeks) than lower gestational age (p = 0.005). Differences in Apgar scores did not reach statistical significance. CONCLUSIONS: The students of the University of the West Indies were more likely to come from smaller families with features indicative of a better quality of life. They were also of higher birthweight and tended to be full-term. The lack of association of Apgar scores with educational attainment is noteworthy.

Small victories, new challenges: two decades of maternal mortality surveillance in Jamaica.

OBJECTIVES: The paper summarizes the changing epidemiology of maternal mortality and the new challenges as Jamaica seeks to contribute to the international goal to reduce maternal mortality by 75% worldwide between 1990 and 2015. METHODS: This is a review of Jamaica's two decades of maternal mortality surveillance experience. RESULTS: Jamaica began episodic reproductive age mortality surveys between 1981 and 1983. In order to move to continuous surveillance, maternal deaths were made a Class 1 notifiable event in 1998. Reporting has steadily improved with over 80% of deaths notified, however events in the first trimester and after the first week post-partum were less likely to be reported. While gestational hypertension remains the leading cause of death, the cause-specific mortality rate has declined in response to specific efforts to reduce its prevalence and consequences. Haemorrhage and infection also declined significantly in prevalence and rank. HIV disease moved rapidly to become the fourth-ranked cause of maternal death. Another lifestyle problem is the growing prevalence of obesity, with its contribution to deaths from heart disease, diabetes mellitus and gestational hypertension. A national maternal mortality surveillance committee has been established to monitor national trends and address policy issues. One of its first tasks will be to develop clinical guidelines to standardize management of the leading direct and indirect complications of pregnancy. CONCLUSION: Jamaica has the capacity to achieve MDG5, however local and national maternal mortality committees and health teams will have to systematically address the deficiencies identified.

Adolescent ears: an avenue into their sexual and reproductive health values.

OBJECTIVES: This study sought to understand what sexual and reproductive health messages Jamaican adolescents get via Dancehall music and how themes in these messages can inform the development of a questionnaire for further exploration of the subject. METHOD: Qualitative processes: key informant interviews, content analysis and focus group discussions were used to identify the themes heard in Dancehall songs to which adolescents listen. Adolescent investigators were utilized to enhance the communication flow among their peers while the adult investigator coordinated the analysis process. The data generated by each method were analysed manually and the themes used to inform development of a quantitative questionnaire for further study. RESULTS: Of 43 songs identified by key informants, twenty-five were short-listed as Dancehall songs as they having met the criteria for conveying sexual and reproductive health and relationship themes. Most themes were gender specific with male specific messages relating to sexual roles and behaviours, physical sexual attributes and performance and sexuality related violence. Female specific messages were about female behaviours in a sexual relationship, financial stability and independence in relationships and physical sexual attributes. All themes were confirmed" through focus group discussions and additional information gleaned about how adolescents perceived the meanings and context of some expressions in the songs analysed. These themes informed the development of a questionnaire in both language and content. CONCLUSIONS: Gender specific issues about sexual relationships and performance, physical sexual attributes and sexual violence were common themes identified in Dancehall songs listed by adolescents. Focus group discussions confirmed adolescents' listenership to the themes and provided explanation of the perceived context and meaning of some messages.

Competition for adolescents' sexual and reproductive health values: is the media winning?

OBJECTIVE: This study aimed to understand the sources and content of sexual and reproductive health (SRH) information among Jamaican adolescents. SUBJECTS AND METHODS: A national survey of adolescents 9-18 years old attending public schools in Jamaica was done using multi-stage stratified cluster sampling techniques. Questions included in a 57-item questionnaire assessed adolescents' awareness of SRH messages promoted by the Ministry of Health (SSRHM) and through Dancehall, (DSRHM). The data were analysed using the Statistical Package for the Social Sciences (SPSS) Version 12.0. Descriptive statistics were used to summarize the data and relationships between variables tested using bivariate: analysis, with a 'p' value of 0.05 accepted as significant. RESULTS: The mean age of respondents was 12.8 (+/- 2.06) years for boys and 12.8 years (+/- 2.23) for girls. Television (76.9%), radio (55.4%) and guidance counsellors (55.2%) were the most common sources of SSRHM with no statistically significant demographic differences. Condom use was the only SSRHM, among the seven assessed, that was reported by more than half of the respondents (85.4% boys and 80.3% girls, p = 0.025). Half the number of the male specific DSRHM assessed was reported by more than 50% of boys. Hurting men who have sex with men, and having many girlfriends were the leading messages reported (69.8% & 65.3%). Among girls, five of nine female-related DSRHM were reported by most girls. Having relationships with males who have material resources (66.8%) and being independent (64%) were the leading messages reported. CONCLUSIONS: The electronic media was the leading source of SSRH messages and condom use was the only SSRHM that outstripped the variety of gender specific DSRH messages, some of which contradict standard messages.

Health status and health maintenance practices among doctors and nurses at two hospitals in Jamaica.

The health of doctors and nurses is of paramount importance because they must be well to perform their jobs optimally under difficult conditions. However, the challenge of their working environment and the culture of their professions often lead to physical and mental illnesses. Despite this, there are several barriers to doctors and nurses seeking healthcare. In this study, the health status and health maintenance practices of doctors and nurses at two hospitals in Kingston, Jamaica, were assessed. This population was previously reported to have a 27% prevalence of probable mental distress based on the General Health Questionnaire 30 (GHQ30). Two hundred and twelve doctors and nurses were recruited into the study. The reported prevalence of chronic diseases was determined while mental health status was based on the GHQ30, reported signs and symptoms of stress and job satisfaction. Health maintenance practices studied included, health-seeking behaviour willingness to seek counselling, reported source of emotional support and use of alcohol and tobacco as coping strategies. Although, less than 50% of study participants reported that they were satisfied with their job, the mean number of days missed from work in the "last six months" was less than two and a half days. The mean time for "last doctor's visit" for nurses and doctors in the current study were 0.93 and 2.4 years, respectively. Females were more willing to seek medical attention than males. More than 50% reported signs and symptoms of stress and major sources of emotional support were friends (55.7%), followed by spouses (36.0%) and colleagues (12.3%). The prevalence of chronic diseases was less than 1% and alcohol and tobacco did not appear to be major coping strategies. The population appeared to be physically healthy and despite the known prevalence of probable mental distress, doctors and nurses appeared unwilling to seek healthcare. Probable barriers to seeking healthcare included confidentiality issues and the need to appear healthy to colleagues, patients and the community.

Are women ready to do HIV test?

OBJECTIVE: To identify the factors that influence the stage of change with regards to HIV testing in women (16- 45-years-old) in Westmoreland using the trans-theoretical model (TTM) of behaviour change. DESIGN AND METHODS: A structured interview-assisted questionnaire was administered to 372 pregnant and non-pregnant respondents in urban and rural areas of Westmoreland after random selection of four public health facilities. The trans-theoretical model which suggests that behaviour change process moves through five stages from pre-contemplation to maintenance was used to evaluate readiness for HIV testing. RESULTS: Most pregnant women who tested previously were at the preparation stage (78.5%) while non-pregnant women who tested previously were at contemplation (68.5%). The significant predictors of being in the action or maintenance stage among pregnant women was being 20- 24-years-old, experiencing a first pregnancy and being exposed to counselling. For women who had never tested, preparation was significantly associated with being in an unstable union (non-pregnant). No significant association was found for non-pregnant, previously tested females or for pregnant women who had never tested. CONCLUSION: The majority of women lacked self-efficacy as they were unable to maintain the behaviour and did not recognize its importance in the absence of pregnancy. Interventions are needed to target non-pregnant women, especially teenagers, women over 25-years old and women in unions. Integration of testing services into all aspects of primary healthcare, established testing protocols and simultaneous marketing to selected at-risk groups will increase the uptake of HIV testing opportunities and contribute to the control of this epidemic.

Maternal mortality surveillance in Jamaica.

OBJECTIVES: To assess factors associated with under-reporting of maternal deaths from 1998, when maternal deaths became a Class I notifiable event in Jamaica and continuous maternal mortality surveillance was introduced, through 2003. METHODS: The number of deaths notified was compared with the number of independently identified deaths for each period and region studied, and key informants reported on their experience of the surveillance process. RESULTS: By 2000, approximately 80% of maternal deaths were reported, and was more consistent in 2 of the 4 regions. In these 2 regions someone was responsible for active surveillance and there was an established maternal mortality committee to review cases. Factors associated with nonreporting were no postmortem examination, death in the first trimester of pregnancy, and time interval between pregnancy termination and death. The surveillance staff requested guidelines on monitoring interregional transfers and technical assistance in developing action plans. CONCLUSION: Active hospital surveillance must include all wards, including the emergency department. Community surveillance should include forensic pathologists. National leadership is needed to summarize trends, address policy, and provide technical assistance to the surveillance staff.

Can research accelerate progress toward millennium development goal 5 (maternal health) in Jamaica?

The Ministry of Health and the University of the West Indies have collaborated over 25 years to develop the evidence base to improve maternal health. The experience is reviewed as a model to accelerate Jamaica's progress toward achieving the Millennium Development Goals (MDGs). The process included measuring the disease burden due to maternal morbidity and mortality; developing and field testing interventions to manage the leading problems, national scale-up, monitoring and evaluation. This began with developing clinical guidelines to manage the hypertensive disorders of pregnancy while establishing high risk (referral) antenatal clinics, expansion and upgrading of referral facilities, and audits to identify barriers to quality healthcare, including establishing maternal mortality surveillance. As we succeed, research funds have become scarce, limiting support to postgraduate students, a reliable, cost effective resource pool capable of undertaking the research needed to provide the evidence base to influence public policy more widely. A locally financed resource pool is needed to support fellowships for graduate students to accelerate their training and availability to contribute to national development. The model from Thailand is put forward for consideration. The operations research model for maternal health can be transferred to other MDG objectives. As Jamaica pursues its goal of developed nation status, and international grant financing shrinks, local civil society will need to fill the vacancy and invest in the most abundant natural resource, young people.

Integrating research into policy and programmes. Examples from the Jamaican experience.

Research into selected health problems across the life cycle are discussed with respect to the application and impact of the findings on policy, programme development and health outcomes. Special emphasis is placed on health research that focussed on the perinatal period, the young child and adolescent, the epidemics of chronic diseases and violence and the linkage between health and tourism. The lessons learnt over more than two decades of practice in the field of public health from conducting research aimed at developing an indigenous evidence base for policies and programmes in Jamaica are summarized.

Timeliness of care for eclampsia and pre-eclampsia in Benin, Ecuador, and Jamaica.

OBJECTIVE: Cases meeting diagnostic criteria for severe pre-eclampsia or eclampsia were reviewed in three countries to determine timeliness and effectiveness of care. METHOD: Cases were retrospectively selected from 11 emergency obstetric care facilities and medical records reviewed by trained obstetricians. RESULT: Of 91 cases (Benin, 28; Ecuador, 25; Jamaica, 38), 74% were correctly treated with anticonvulsant and 77% with antihypertensive therapy. The median interval to treat eclampsia (anticonvulsant, 28 min; antihypertensive, 77 min) was shorter than for severe pre-eclampsia (anticonvulsant, 45 min; antihypertensive, 85 min). Two in three cases (65%) received anticonvulsant but only 41% received antihypertensive therapy within 60 min of diagnosis. While 74% of eclamptics had been delivered within 12 h, only 39% of severe pre-eclamptics were delivered within 24 h. CONCLUSION: Timeliness can be studied in developing countries. Its objective measurement is a first step towards improving this component of care.

Epidemiologic transition in maternal mortality and morbidity: new challenges for Jamaica.

OBJECTIVE: Given interventions implemented in recent years to reduce maternal deaths, we sought to determine the incidence and causes of maternal deaths for 1998-2003. METHOD: Records of public hospitals and state pathologists were reviewed to identify pregnancy-related deaths within 12 months of delivery and determine their underlying causes. RESULTS: Maternal mortality declined (p=0.023) since surveillance began in 1981-83. The fall in direct mortality (p=0.0003) included 24% fewer hypertension deaths (introduction of clinical guidelines, reorganization of antenatal services) and 36% fewer hemorrhage deaths (introduction of plasma expanders). These improvements were tempered by growing indirect mortality (p=0.057), moving to 31% of maternal deaths from 17% in 1993-95. INTERPRETATION: Declines in direct mortality may be associated with surveillance and related improvements in obstetric care. Increased indirect deaths from HIV/AIDS, cardiac disease, sickle cell disease and asthma suggests the need to improve collaboration with medical teams to implement guidelines to care for pregnant women with chronic diseases.

Hurricane preparedness among healthcare workers in St James.

OBJECTIVE: To assess the preparedness of health workers in St James, Jamaica, to respond to natural disasters. METHODS: A cross-sectional survey was conducted on a 25% quota sample of hospital and health department staff in St James in 2005 (n = 307). Awareness of and attitudes to disaster management policies, plans and training and how these influenced their response in Hurricane Ivan were evaluated Statistical package for the Social Science 11.5 was used to summarize quantitative data, while qualitative data were analyzed manually. RESULTS: Most respondents (67%) knew of the disaster plan but only 40% had been trained in disaster management. More nurses (68%) and paramedicals (51%) reported being trained than ancillary/ auxilliary (33%), medical (21%) or administrative/clerical (18%) staff Most (96%) had participated in at least one disaster preparedness drill, usually a fire drill, but not in the previous two years. Attitudes towards disaster management were positive (99% insisted training should be compulsory, 95% agreed they should help in a hurricane). The majority (86%) reported being available to work at any time and 77% had worked as scheduled during Hurricane Ivan. Transportation and personal responsibility (to children, elderly or property) were the main limiting factors. Provisions for staff welfare (eg transportation, protective gear, media updates) were perceived as inadequate. Having defined disaster response roles was positively correlated with age, years of service and occupation (p < 0.01) CONCLUSION: Health workers in St James have positive attitudes to disaster management but require routine training to compensate for staff turnover. More attention must be given to staff welfare during disasters.

Mental well-being of doctors and nurses in two hospitals in Kingston, Jamaica.

Hospital work involves some of the most stressful situations found in any workplace. Furthermore, hospital workers may be affected by non-work-related stress such as family responsibilities and financial difficulties, leading to impaired mental well-being and suboptimal performance. The aim of this study was to assess the level of general mental well-being among doctors and nurses from two hospitals in Kingston, Jamaica. A total of 212 doctors and nurses at the Kingston Public Hospital and the University Hospital of the West Indies were studied yielding a participation rate of 83.1%. A self-administered questionnaire was used to gather social and biomedical data and the General Health Questionnaire 30 (GHQ 30) used to determine general mental well-being. Probable caseness was defined as a GHQ 30 score > 5. Focus group discussions were also held with staff at both hospitals. A total of 27.4% of the study population met the GHQ-30 criteria (caseness) defining them as probable cases of mental distress. Cases and non-cases were not different in age, gender or hospital of employment. However, caseness was associated with years of professional experience, work-related and non-work-related stress, serious financial difficulties and fears of coming to work. Significant predictors of increased risk of caseness were fear of coming to work (OR 3.06; CI 1.40, 6.70); professional experience in excess of five-years and high non-work-related stress. High work-related stress was associated with reduced risk of being classified a case, suggesting that work may have been therapeutic. Focus group discussions suggested that non-work stress was related to financial difficulties, commuting and child care, especially among nurses. Intervention to improve general mental well-being should be targeted at new employees and should address child care, commuting and financial management.

Growth curves for normal Jamaican neonates.

The aim of this study was to provide standards for the assessment of birthweight, head circumference and crown-heel length for normal, singleton newborns of predominantly West African descent. Data were collected for 10 482 or 94% of all recorded births in Jamaica during the two-month period September 1 to October 31, 1986. After editing procedures, data were available for 6178 (birthweight), 5975 (head circumference), and 5990 (crown-heel length). The data presented in tables and growth curves include birthweight, head circumference and crown-heel length for males and females separately, for gestational ages 30-43 weeks. Data sets from the University Hospital of the West Indies in 1990 and 1999 were used to explore the possibility of secular change over the period 1986-1999. In conclusion, these ethnic and gender-specific growth curves are based on the most extensive dataset currently available in Jamaica for babies of West African descent.

Registration of births, stillbirths and infant deaths in Jamaica.

BACKGROUND: Vital statistics underestimate the prevalence of perinatal and infant deaths. This is particularly significant when these parameters affect eligibility for international assistance for newly emerging nations. OBJECTIVE: To determine the level of registration of livebirths, stillbirths and infant deaths in Jamaica. METHODOLOGY: Births, stillbirths and neonatal deaths identified during a cross-sectional study (1986); and infant deaths identified in six parishes (1993) were matched to vital registration documents filed with the Registrar General. RESULTS: While 94% of livebirths were registered by one year of age (1986), only 13% of stillbirths (1986) and 25% of infant deaths (1993) were registered. Post neonatal deaths were more likely to be registered than early neonatal deaths. Frequently the birth was not registered when the infant died. Birth registration rates were highest in parishes with high rates of hospital deliveries (rs = 0.97, P < 0.001) where institutions notify the registrar of each birth. Hospital deaths, however, were less likely to be registered than community deaths as registrars are not automatically notified of these deaths. CONCLUSIONS: To improve vital registration, institutions should become registration centres for all vital events occurring there (births, stillbirths, deaths). Recommendations aimed at modernizing the vital registration system in Jamaica and other developing countries are also made.

PIP: Vital statistics indicate only part of the actual prevalence of perinatal and infant mortality. Findings are reported from a study conducted to determine the level of registration of live births, stillbirths, and infant deaths in Jamaica. Births, stillbirths, and neonatal deaths identified during a 1986 cross-sectional study and infant deaths identified in six parishes during 1993 were matched to vital registration documents filed with the Registrar General. While 94% of live births were registered by one year of age, only 13% of stillbirths and 25% of infant deaths were so registered. Post neonatal deaths were more likely to be registered than early neonatal deaths. Frequently the birth was not registered when the infant died. Birth registration rates were highest in parishes with high rates of hospital deliveries where institutions notify the registrar of each birth. Hospital deaths, however, were less likely to be registered than community deaths since registrars are not automatically noticed of such deaths. Institutions should register all vital events occurring there.

Access to care and maternal mortality in Jamaican hospitals: 1993-1995.

BACKGROUND: As part of the reproductive health quality assurance programme, the Ministry of Health sought to review maternal deaths in public hospitals. These hospitals attend 95% of institutional births and 82% of all births. METHODS: Deaths among females 10-50 years in public hospitals during 1993-1995 were reviewed to identify pregnancy-related deaths. Cause of death and access to care were compared with previous studies (1981-1983 and 1986-1987 [12 months]). RESULTS: The maternal mortality ratio of 106.2 per 100 000 live births, was no different than the 119.7 observed in 1986-1987 and 118.6 for 1981-1983. The leading causes of death remained pre-eclampsia/eclampsia and haemorrhage. The only significant cause-specific decline occurred among deaths due to ruptured ectopic pregnancy (P = 0.012). While in 1986-1987 access to care was associated with risk of death from gestational hypertension (P = 0.02), these differences are no longer significant. Differences persist, however, for haemorrhage and all other causes, which were less likely to occur at the more skilled institutions. The region with the least obstetricians had the highest mortality ratio but the one with the most did not have the lowest ratio, indicating that quality is more important than quantity. CONCLUSIONS: Regional differences indicate the capacity to reduce maternal mortality by at least 50% with re-allocation of skilled personnel and improved quality. All hospitals must be able to manage haemorrhage cases as patients are unlikely to survive referral.