Neonatal mortality, risk factors and causes: a prospective population-based cohort study in urban Pakistan

Objective

To evaluate the prevalence, sex distribution and causes of neonatal mortality, as well as its risk factors, in an urban Pakistani population with access to obstetric and neonatal care.

Methods

Study area women were enrolled at 20–26 weeks’ gestation in a prospective population-based cohort study that was conducted from 2003 to 2005. Physical examinations, antenatal laboratory tests and anthropometric measures were performed, and gestational age was determined by ultrasound to confirm eligibility. Demographic and health data were also collected on pretested study forms by trained female research staff. The women and neonates were seen again within 48 hours postpartum and at day 28 after the birth. All neonatal deaths were reviewed using the Pattinson et al. system to assign obstetric and final causes of death; the circumstances of the death were determined by asking the mother or family and by reviewing hospital records. Frequencies and rates were calculated, and 95% confidence intervals were determined for mortality rates. Relative risks were calculated to evaluate the associations between potential risk factors and neonatal death. Logistic regression models were used to compute adjusted odds ratios.

Findings

Birth outcomes were ascertained for 1280 (94%) of the 1369 women enrolled. The 28-day neonatal mortality rate was 47.3 per 1000 live births. Preterm birth, Caesarean section and intrapartum complications were associated with neonatal death. Some 45% of the deaths occurred within 48 hours and 73% within the first week. The primary obstetric causes of death were preterm labour (34%) and intrapartum asphyxia (21%). Final causes were classified as immaturity-related (26%), birth asphyxia or hypoxia (26%) and infection (23%). Neither delivery in a health facility nor by health professionals was associated with fewer neonatal deaths. The Caesarean section rate was 19%. Almost all (88%) neonates who died received treatment and 75% died in the hospital.

Conclusion

In an urban population with good access to professional care, we found a high neonatal mortality rate, often due to preventable conditions. These results suggest that, to decrease neonatal mortality, improved health service quality is crucial.     

Blood pressure and salt intake in Malawi: an urban rural study.

A significant difference between the blood pressures of rural and urban Malawians was found in both sexes, was present at the age of 15 years, and was associated with obesity but not with smoking, alcohol consumption, occupation or housing. Pulse rate was significantly lower in the urban group. These differences were accompanied by low potassium and sodium intake although the sodium intake in the urban group was double that in the rural group. No direct relation between blood pressure and urinary electrolytes was found.     

Retinopathy of prematurity: risk factors in a prospective population-based study

A prospective study of risk factors for retinopathy of prematurity (ROP) in all very low birthweight (less than 1500 g) infants born in New Zealand in 1986 is reported. Of 413 liveborn infants admitted to neonatal units, 338 (81.2%) survived to be discharged home. Of surviving infants, 313 (93%) were examined by indirect ophthalmoscopy, as were eight infants who died before discharge. Sixty-nine infants (21.5% of 321) had acute retinopathy. On multiple logistic regression analysis, three variables made statistically significant independent contributions to the risk of any acute retinopathy; gestational age (P less than 0.0001), principal hospital caring for the infant (P less than 0.01) and treatment with indomethacin (P less than 0.01). Only two variables, gestational age (P less than 0.0001) and hospital (P less than 0.01), made significant contributions to the risk of stage 2 or more ROP. For both categories of ROP, timing of the examination also had a statistically significant effect (P less than 0.001). After adjustment for other significant predictor variables, it was estimated that approximately 70% of infants of less than 26 weeks' gestation were at risk of ROP and nearly 50% of stage 2 or more ROP, in comparison with less than 2% of infants of 32 weeks' gestation or more; infants treated with indomethacin were over 1.5 times more likely to have ROP than infants not so treated. Failure to enforce uniform timing of examination was the most serious defect in the study; only 205 (64%) of the 321 infants were examined at the recommended time. However, reanalysis of the model with information limited to these 205 infants yielded similar risk factors. The incidence of ROP, both observed (P less than 0.05) and adjusted for other significant variables in the regression model (P less than 0.01) was lowest in the two largest level III hospitals. These hospitals also had the best survival rates after adjustment for birthweight, gestation and gender (P less than 0.01). We speculate that the larger level III units obtained better results because their size and experience enabled them to provide a better overall quality of care.     

Prevalence and risk factors of non-insulin-dependent diabetes mellitus in the rural and urban population of Tunisia

BACKGROUND: Prevalence of non-insulin-dependent diabetes mellitus was analyzed from a baseline survey of risk factors for cardiovascular disease in a random sample of the 35-50-year old population of the Gouvernorat of Nabeul, Tunisia. METHODS: Diabetes was assessed on the basis of an interview and fasting blood glucose level among a sample of 692 men and women (35% aged 35-40 years, 30% aged 40-45 years, 35% aged 45-50 years). RESI;TS: Forty-six percent of the sample were men. The prevalence of diabetes was 7.2%; two-thirds of the identified cases of diabetes were known. The prevalence of diabetes was much higher in the urban area (9.3% in men and 10.4% in women) compared with the rural area (2.0% in men and 4.5% in women). Mean body mass index was significantly different (p<0.01) between normal and diabetic subjects. Hypertension was three times higher in diabetic subjects (33%) than in normal subjects (9%). Triglycerides level was higher in diabetic subjects (1.92+/-1.72mmol/l) compared with non-diabetic subjects (1.29+/-1.02mmol/l). There was a positive relation between blood glucose level and triglycerides level, independently of obesity in women, but dependently in men. CONCLUSION: In the Tunisian population, known to have a low level of cardiovascular risk factors, the relationships between diabetes, obesity, hypertension, and lipid abnormalities are similar to those observed in Western populations.     

Diabetes risk factors, diabetes and diabetes care in a rural Australian community

OBJECTIVE: To comprehensively describe diabetes-related risk factors, quality of care and patient-perceived barriers to care in a rural community. DESIGN: Cross-sectional mail study, self-completed survey and retrospective chart review. SETTING: Community and health services in Corryong, rural Victoria, Australia. PARTICIPANTS: Ninety-seven patients with diabetes and 495 with other diseases in the mail study, 84 with diabetes in the self-completed survey and 101 diabetic patient chart reviews. MAIN OUTCOME MEASURES: Self-reported lifestyle activities, uptake of health checks, metabolic measures and uptake of medication, and self-reported barriers to diabetes care. RESULTS: Most residents without diabetes had recently had their blood pressure and cholesterol checked; 60.4% were trying to control their weight and 73.9% were exercising regularly (although only 30.7% to an adequate level). Those with diabetes reported a greater uptake of healthy living messages, and had a mean HbA1c of 7.3%, total cholesterol of 5.0 mmol L(-1); 12.9% had a diastolic blood pressure > or =85 mmHg. Foot checks were infrequent (18%). There was substantial room to increase antiplatelet, blood pressure, antihyperglycaemia and lipid-lowering therapy. Most patients reported psychological (84.5%) and educational (82.1%) barriers to care, with few perceiving physical barriers to care. CONCLUSION: Living in a rural area with predominantly GP care can be associated with comparatively good metabolic control, although psycho-educational barriers are frequently present. In the wider community, risk factors for diabetes remain common, and the majority have been screened for components of the metabolic syndrome in the previous year.     

成都市城乡居民糖尿病患病率及影响因素研究

目的 探讨成都市城乡居民糖尿病患病率及其影响因素。方法 采用整群抽样方法 抽取1847名成都市城乡居民,利用问卷调查居民一般情况及过去一年食物摄人情况,测量身高、体重,并行口服葡萄糖耐量试验,在总人群和城乡居民中分别以因子分析提取膳食模式,并以多因素非条件logistic回归分析糖尿病影响因素。结果 调查对象中糖尿病患病率达23.2%,人口标化患病率为20.2%;城乡居民糖尿病患病率分别为28.7%和11.1%。经logistic回归分析,总人群中,与青年人比较,中年人(OR=2.337,95%Cl:1.305～4.185)和老年人(OR=5.990,95%Cl:3.389～10.586)有更高的糖尿病患病风险;职业为管理人(OR=1.434,95%Cl:1.000～2.057)和普通职员(OR=2.870,95%CI:1.653～4.980)较务农人员有更高的患病风险。城市居民中,中年人(OR=2.973,95%CI:1.101～8.031)和老年人(OR=5.972,95%Cl:2.267～15.730)与青年人比较有更高的患病风险;职业为普通职员(OR=2.196,95%C1:1.213～3.975)较务农人员有更高的患病风险;在农村居民中,高能量高蛋白膳食模式者(OR=1.404,95%Cl:1.113-1.772)有更高的糖尿病风险。结论 成都市城乡居民糖尿病患病率较高,年龄、职业和膳食结构均对糖尿病有影响,且影响因素存在差异,营养改善工作应按照城乡人群特点区分。     

Maternal risk factors for fetal alcohol syndrome in the Western cape province of South Africa: a population-based study

OBJECTIVES: We defined risk factors for fetal alcohol syndrome (FAS) in a region with the highest documented prevalence of FAS in the world. METHODS: We compared mothers of 53 first-grade students with FAS (cases) with 116 randomly selected mothers of first-grade students without FAS (controls). RESULTS: Differences between case and control mothers in our study population existed regarding socioeconomic status, religiosity, education, gravidity, parity, and marital status. Mothers of children with FAS came from alcohol-abusing families in which heavy drinking was almost universal; control mothers drank little to no alcohol. Current and past alcohol use by case mothers was characterized by heavy binge drinking on weekends, with no reduction of use during pregnancy in 87% of the mothers. Twenty percent of control mothers drank during pregnancy, a rate that declined to 12.7% by the third trimester. The percentage who smoked during pregnancy was higher for case mothers than for control mothers (75.5% vs 30.3%), but the number of cigarettes smoked was low among case mothers. The incidence of FAS in offspring of relatively young women (28 years) was not explained by early drinking onset or years of drinking (mean, 7.6 years among case mothers). In addition to traditional FAS risk factors, case mothers were smaller in height, weight, head circumference, and body mass index, all anthropomorphic measures that indicate poor nutrition and second-generation fetal alcohol exposure. CONCLUSIONS: Preventive interventions are needed to address maternal risk factors for FAS.     

Towards a subsiding diabetes epidemic: trends from a large population-based study in Israel

Background

With increasing diabetes prevalence worldwide, an impending diabetes “pandemic” has been reported. However, definitions of incident cases and the population at risk remain varied and ambiguous. This study analyzed trends in mortality and screening that contribute to diabetes prevalence and incidence, distinguishing between new incident cases and newly detected cases.

Methods

In an integrated provider-and-payer-system covering 53% of Israel’s population, a composite diabetes case-finding algorithm was built using diagnoses, lab tests, and antidiabetic medication purchases from the organization’s electronic medical record database. Data were extracted on adult members aged 26+ each year from January 1, 2004 through December 31, 2012. Rates of diabetes prevalence, incidence, screening, and mortality were reported, with incidence rates evaluated among the total, “previously-screened,” and “previously-unscreened” at-risk populations.

Results

There were 343,554 diabetes cases in 2012 (14.4%) out of 2,379,712 members aged 26+. A consistent but decelerating upward trend in diabetes prevalence was observed from 2004-2012. Annual mortality rates among diabetics decreased from 13.8/1000 to 10.7/1000 (p = 0.0002). Total population incidence rates declined from 13.3/1000 in 2006 to 10.8/1000 in 2012 (p < 0.0001), with similar incidence trends (13.2/1000 to 10.2/1000; p = 0.0007) among previously-screened at-risk members, and a rise in testing rates from 53.0% to 66.7% (p = 0.0004). The previously-unscreened group decreased 28.6%, and the incidence rates within this group remained stable.

Conclusions

The increase in diabetes prevalence is decelerating despite declining mortality and increasing testing rates. A decline in previously-screened incident cases and a shrinking pool of previously-unscreened members suggests that diabetes trends in Israel are moving toward equilibrium, rather than a growing epidemic.

     

Type 2 diabetes and impaired fasting blood glucose in rural Bangladesh: a population-based study

BACKGROUND: Diabetes is a fast expanding global health problem but more so in the developing countries. Therefore, it is of particular interest to study the epidemiological transition of the state and to identify the risk factors in order to recognize the extent of the problem. METHODS: A random sample of 5000 rural individuals (age >/=20 years) were included in a cross-sectional study. Fasting capillary blood glucose levels were measured from 4757 individuals. Height, weight, waist, hips including blood pressure and demographic information was collected. RESULTS: The study population was lean [mean body mass index (BMI) 19.4] with a low prevalence of type 2 diabetes but relatively high impaired fasting glucose. No relationship between type 2 diabetes and BMI in men, but an overall relationship was observed for women (P = 0.04) (data not shown). Age, sex, and waist/hip ratio appeared to be important risk factors for the occurrence of type 2 diabetes in this population. CONCLUSIONS: Low prevalence of type 2 diabetes and relative high impaired fasting blood glucose was observed. The factors associated with the occurrence of diabetes in this population appeared to differ than its known relations with BMI. This may indicate that the risk factors for type 2 diabetes are likely to differ in different population. Our results are likely to be in line with the Indian data suggesting that a revised guideline for anthropometric measures in the South Asian population is called for, in order to classify people at risk.     

Infections and risk of type I diabetes in childhood: a population-based case-control study

Objective: This study focuses on the evaluation of some infectious diseases as risk determinants of type I diabetes mellitus (DM). Methods: A population-based case–control study was carried out by referring to the type I DM population-based register of the Abruzzo region of Italy as it includes all type I DM cases since January 1 1990, the point at which the register became operative. The pediatric population (age: 0–14), living in the same municipalities of the cases, was selected as the control population. Data were collected through questionnaires submitted by a physician to parents of cases and controls. Conditional logistic regression models were used to evaluate association between determinants and onset of type I DM. Results: The risk of diabetes for children exposed to only one infection (morbilli, parotitis, rubella, pertussis or varicella) is not statistically significant: OR: 0.778; CI: 0.427–1.370. On the contrary, when two infections are contracted statistically significant results occur: OR: 2.375; CI: 1.149–4.914; for more than two infections values are: OR: 6.786; CI: 2.881–17.877. No substantial difference in odds ratios (ORs) after adjustment for confounding variables was found. A significant decrease in OR was noted for pertussis and MMR vaccinations, respectively: OR: 0.015; CI: 0.001–0.251; OR: 0.400; CI: 0.201–0.799. Conclusions: Since the higher the number of contracted infections, the higher the risk of diabetes, contracted infections can be considered potential accelerating factors of clinical manifestation of type I DM. Therefore multiple exposures might speed up the onset of diabetes in children. This study suggests the utility of applying the risk model method to wider populations, especially if the geographical variability of standardised incidence rates of type I DM in pediatric age is taken into consideration.     

江西省城乡居民高血压流行情况及影响因素研究

目的了解江西省城乡居民高血压流行情况及其影响因素,为制订防治策略提供参考。方法采用多阶段分层随机抽样的方法,抽取江西省18岁及以上的常住居民开展慢性病流行现状调查,分析高血压患病流行情况及相关影响因素。结果高血压患病率为30.67%,标化率24.46%;男性高于女性(χ2=4.23,P<0.05),差异有统计学意义;城市低于农村(χ2=8.00,P<0.05),差异有统计学意义;增龄(OR=2.361)、体质指数高(OR=1.170)、喜食咸菜(OR=1.089)、胆固醇高(OR=1.764)为高血压危险因子,文化程度高(OR=0.783)为保护因子。控制增龄因素后,体质指数、咸菜摄入、胆固醇、糖化血红蛋白、血糖与收缩压有一定的线性相关,差异均有统计学意义;收缩压、舒张压与文化程度Spearmen等级相关系数分别为-0.27和-0.11,差异有统计学意义。结论应加强对江西省居民尤其是农村人群的血压、血糖和体重监测,开展患者及高危人群健康教育与干预。     

Neurodevelopmental outcome and risk factors for impaired development of African American infants in an underserved urban population: a population-based study

We aimed 1) to define risk factors for adverse outcome in urban African American patients, 2) to determine whether clinical variables as risk factors are congruent with previously published data, and 3) to identify the proportion of infants with different outcomes. The study included African American infants who were born and participated in neurodevelopmental follow-up. Infants with gestational age range of 23 to 41 weeks, and birth weight (BW) range of 495 to 3,965 g were classified by developmental outcome. Among the smallest infants, BW, gestational age, gender and respiratory distress syndrome were significantly (p<.05) associated with adverse outcome. No significant risk factors were identified for adverse outcome in the two other birth weight categories. Adverse outcomes were seen more frequently in infants with BW ≤1,500 g than in larger infants. The number of infants with severe outcome was found higher than previously reported and may be related to different racial/generational origin.     

农村居民糖尿病患病率及危险因素分析

目的 了解农村居民的糖尿病患病率及其危险因素,为评价农村居民糖尿病的流行现状和采取相应的预防措施提供依据.方法 在河南省选择2个村,随机抽取18岁及以上的居民进行问卷调查、体格检查和血液生化指标检测,描述糖尿病的患病率分布,并应用Logistic回归模型分析其危险因素.结果 共调查989人,糖尿病(DM)患病率为11.22%,其中男性为10.36%,女性为6.88%;用2000年全国人口普查数据进行年龄标化后患病率为6.72%.DM患病率在男女之间差异无统计学意义;DM患病率随着年龄的增加而增高,随着收入水平的增高而增高;低文化水平人群的DM患病率高于高文化水平人群.调整性别、年龄、婚姻状况、文化水平和经济收入后,糖尿病的独立危险因素是中心性肥胖(OR=2.066,OR95%CI=1.298~3.289)、糖尿病家族史(OR=2.981,OR95%CI=1.687~5.266)和低密度脂蛋白(LDL-C)≥3.37(OR=1.601,OR95%CI=1.019~2.518).结论 农村居民糖尿病患病率已经达到较高的水平,成为农村地区重要的公共卫生问题,亟需加强农村居民的糖尿病预防控制工作.     

The epidemiology of premenstrual symptoms in a population-based sample of 2650 urban women: attributable risk and risk factors.

This epidemiologic survey achieved a reliable measure of the prevalence of premenstrual symptoms by avoiding the biases of small or selected samples, anamnestic error, and subjective expectation. From 6232 women (a 78.8% response), aged 20-49 years, identified through a random sample of urban households, the 24-hour prevalence of symptoms was obtained using the Moos' Menstrual Distress Questionnaire, administered without reference to the menstrual cycle. For 71% of the naturally cycling women, current cycle phase was determined by follow-up (n = 2650); but a higher prevalence of severe or moderately severe affective symptoms in the premenstrual compared to the mid-cycle phase was not found. However, observed risk factor interactions led to the conclusion that premenstrual distress is a discrete mood disorder, affecting women aged 25-35 years, with probable ovulatory cycles, and vulnerable to stress; and that the risk of affective symptoms attributable to the premenstrual state was one percent.     

Use of medication by the elderly in urban and rural areas in southern Brazil: a population-based study

The study aimed to measure use of medication and polypharmacy among the elderly in Carlos Barbosa, Rio Grande do Sul State, Brazil, and to compare socio-demographic, economic, and health characteristics in relation to area of residence (urban versus rural) in a random sample of 811 persons 60 year of age or older. Interviews were used to collect data on socio-demographic characteristics, chronic illnesses, and self-reported use of medications. The association between area of residence and medication or polypharmacy was adjusted for confounders using Poisson regression with robust variance. Prevalence rates for use of medication and polypharmacy were higher among older persons living in the urban area. Living in the urban area was positively and independently associated with use of medication (PR = 1.10; 95%CI: 1.02-1.20) and polypharmacy (PR = 1.83; 95%CI: 1.27-2.65) in this group of elderly in southern Brazil.     

农村贫困地区脑卒中危险因素及经济负担的病例对照研究

目的 研究农村贫困地区脑卒中危险因素及经济负担。方法 采用病例对照研究，在辽宁省彰武县农村选取脑卒中患者和对照农民各102人，进行问卷调查。结果 脑卒中的危险因素在彰武县农村农民中流行率较高，吸烟、饮酒、喝茶的比例分别达73．1％、64．7％、79．4％以上。很少吃水果、蛋类和动物性蛋白质者分别为61．8％、56．9％、61．8％以上。每人每天食盐和猪油摄入量分别为22．2g和31．3g，具有高血压和高血脂患病意识者仅为19．6％和20．6％。脑卒中患者中高血压家族史、饮酒、超重、低动物蛋白摄人率分别为31．4％、76．5％、23．5％和75．5％，显著高于对照组。脑卒中组人均年纯收人为。704．9元，对照组为959．4元，脑卒中患者医疗费用年平均为3120．6元。最多者每年医疗费用高达39000元，远远超出了其家庭纯收入。结论 高血压、饮酒、超重、低蛋白摄人、吸烟、高盐和高脂肪摄入可能是脑卒中的危险因素，脑卒中患病使患者家庭背负了沉重的经济负担。     

Occupational risk factors for brain cancer: a population-based case-control study in Iowa

A number of occupations and industries have been inconsistently associated with the risk of brain cancer. To further explore possible relationships, we conducted a population-based case-control study of brain glioma in the state of Iowa, involving 375 histologically confirmed incident cases and 2434 population-based controls. Among men, the industries and/or occupations that had a significantly increased risk for employment of more than 10 years included roofing, siding, and sheet metalworking; newspaper work; rubber and plastics products, particularly tires and inner tubes; miscellaneous manufacturing industries; wholesale trade of durable goods, grain, and field beans; cleaning and building service occupations; miscellaneous mechanics and repairers; and janitors and cleaners. Subjects who worked in plumbing, heating, and air conditioning; electrical services; gasoline service stations; and military occupations also experienced a significantly increased risk. Among women, significant excess risk was observed for occupations in agricultural services and farming, apparel and textile products, electrical and electronic equipment manufacturing, various retail sales, record-keeping, and restaurant service. Workers in industries with a potential for gasoline or motor exhaust exposures experienced a non-significant excess risk of brain glioma.     

Psychosocial risk factors for the onset of abdominal pain. Results from a large prospective population-based study

OBJECTIVE: To determine the psychosocial risk factors for the development of abdominal pain and to determine whether, in those people who consulted, symptoms had been attributed to an organic cause. DESIGN: Prospective population-based postal survey with follow-up survey at 12 months. SETTING: A mixed sociodemographic suburban area of Manchester, UK. PARTICIPANTS: Subjects aged 18-65 years were randomly selected from a population-based primary care register who had responded to a detailed pain questionnaire, which included a pain manikin drawing. They also completed the following psychosocial instruments: General Health Questionnaire, Somatic Symptom Checklist, Fatigue Questionnaire and the Illness Attitude Scales (including the 'health anxiety' and 'illness behaviour' sub-scales). MAIN OUTCOME MEASURES: The onset of new abdominal pain. RESULTS: Of the 1953 participants at baseline, 1763 were free of abdominal pain: 1551 were followed up at 12 months (adjusted follow-up rate of 92%) of which 69 subjects reported new abdominal pain (new onset rate 4.6%). New abdominal pain was similar in females (4.9%) and males (4.2%), and did not vary by age group. Baseline factors which predicted onset were high levels of fatigue (odds ratio [OR] = 3.3, 95% CI: 1.9-5.8), psychological distress (OR = 3.4, 95% CI: 1.9-6.0), high scores on the illness behaviour scale (OR = 3.3, 95% CI: 1.7-6.7) and high levels of health anxiety (OR = 2.1, 95% CI: 1.1-3.9). Reporting low back pain at baseline was also associated with an increased risk of reporting abdominal pain (OR = 2.0, 95% CI: 1.2-3.3). On multivariate analysis, high levels of psychological distress and aspects of prior illness behaviour were the major independent predictors of outcome. Of those who sought health care, only one consultation led to a definite diagnosis. CONCLUSION: In subjects free of abdominal pain, psychological distress, fatigue, health anxiety and illness behaviour are predictors of future onset rather than merely a consequence of symptoms. These results suggest that abdominal pain shares some common features of onset with pain at other sites thought not to be primarily organic in origin.     

Understanding enrolment in community health insurance in sub-Saharan Africa: a population-based case-control study in rural Burkina Faso

OBJECTIVE: To identify factors associated with decision to enrol in a community health insurance (CHI) scheme. METHODS: We conducted a population-based case-control study among 15 communities offered insurance in 2004 in rural Burkina Faso. For inclusion in the study, we selected all 154 enrolled (cases) and a random sample of 393 non-enrolled (controls) households. We used unconditional logistic regression (applying Huber-White correction to account for clustering at the community level) to explore the association between enrolment status and a set of household head, household and community characteristics. FINDINGS: Multivariate analysis revealed that enrolment in CHI was associated with Bwaba ethnicity, higher education, higher socioeconomic status, a negative perception of the adequacy of traditional care, a higher proportion of children living within the household, greater distance from the health facility, and a lower level of socioeconomic inequality within the community, but not with household health status or previous household health service utilization. CONCLUSION: Our study provides evidence that the decision to enrol in CHI is shaped by a combination of household head, household, and community factors. Policies aimed at enhancing enrolment ought to act at all three levels. On the basis of our findings, we discuss specific policy recommendations and highlight areas for further research.