A probabilistic method to estimate the burden of maternal morbidity in resource-poor settings: preliminary development and evaluation
© Fottrell et al.; licensee BioMed Central Ltd. 2014
Received: 12 July 2013
Accepted: 18 February 2014
Published: 13 March 2014
Maternal morbidity is more common than maternal death, and population-based estimates of the burden of maternal morbidity could provide important indicators for monitoring trends, priority setting and evaluating the health impact of interventions. Methods based on lay reporting of obstetric events have been shown to lack specificity and there is a need for new approaches to measure the population burden of maternal morbidity. A computer-based probabilistic tool was developed to estimate the likelihood of maternal morbidity and its causes based on self-reported symptoms and pregnancy/delivery experiences. Development involved the use of training datasets of signs, symptoms and causes of morbidity from 1734 facility-based deliveries in Benin and Burkina Faso, as well as expert review. Preliminary evaluation of the method compared the burden of maternal morbidity and specific causes from the probabilistic tool with clinical classifications of 489 recently-delivered women from Benin, Bangladesh and India.
Using training datasets, it was possible to create a probabilistic tool that handled uncertainty of women’s self reports of pregnancy and delivery experiences in a unique way to estimate population-level burdens of maternal morbidity and specific causes that compared well with clinical classifications of the same data. When applied to test datasets, the method overestimated the burden of morbidity compared with clinical review, although possible conceptual and methodological reasons for this were identified.
The probabilistic method shows promise and may offer opportunities for standardised measurement of maternal morbidity that allows for the uncertainty of women’s self-reported symptoms in retrospective interviews. However, important discrepancies with clinical classifications were observed and the method requires further development, refinement and evaluation in a range of settings.
KeywordsMaternal health Morbidity Developing countries Pregnancy Childbirth Bayesian analysis Africa Asia
The aim of most safe motherhood programmes in resource-poor settings is to reduce maternal mortality and morbidity. There is great interest from funders, policy makers and researchers in evaluating their success using health outcomes, particularly a reduction in deaths or severe complications. Measurement of maternal mortality in these settings is notoriously elusive, however, given its relative rarity, the large sample sizes needed and the reliance on verbal autopsy methods to identify pregnancy status and causes of death when it occurs at home . For every maternal death there are a large number of women who suffer illness and may come close to death and suffer long-term consequences of obstetric morbidity [2–4]. Population-based estimates of the burden of maternal morbidity, therefore, could be useful indicators for monitoring trends, priority setting and evaluating the health impact of interventions [5, 6], particularly within a context of falling maternal mortality. Measuring the burden of maternal morbidity is also difficult, however, particularly in populations where many women deliver at home and may go through pregnancy and the post-partum period with limited contact with health services.
Data on a sample of hospital users are unlikely to provide a representative picture of all maternal morbidity at the population level, although new methods do show promise for extreme, life-threatening conditions . Furthermore, women’s ability to accurately recall and report signs or symptoms related to diagnoses of complications is limited, mostly because of lack of specificity, thereby leading to difficulties in estimating prevalence [8–11]. Thus, retrospective interviews on women’s perceived obstetric complications tend to over-estimate the burden and evaluations of various approaches have generally shown poor validity .
In general, community-based survey approaches have relied on reports of the presence or absence of specific signs and symptoms in order to ascertain a definitive binary outcome (whether or not the woman experienced a specific complication). Clinical review of the data, with or without decision tree algorithms, may be used to classify cases into specific morbidity cause categories . However, such approaches do not generally account for uncertainty of lay recall and reporting of signs, and the derived morbidity diagnoses based on binary classifications may falsely imply certainty of classification. Forced dichotomy of the outcome based on uncertain symptom histories may partly explain the over-estimates of morbidity based on these methods.
Identifying multiple possible complications and their causes with specified degrees of likelihood, which may then be aggregated to provide a profile of cause-specific morbidity burdens at the population level, may be a more realistic endeavour than seeking crude binary classifications based on self-reports of questionable validity. This paper describes the development and preliminary evaluation of an innovative probabilistic approach to handling women’s self-reports of pregnancy and delivery experiences, signs and symptoms, to estimate population-level burdens of obstetric morbidity and its causes as needed by local health managers and researchers in resource-and data-poor settings.
Theory & technical overview
List of the 72 signs and symptoms (collectively called 'indicators’) and the 10 direct and indirect causes of obstetric complications included in the InterSAMM probabilistic model
1. aged under 20 yrs
25. any diagnosis of anaemia
49. did she visit more than one health facility
2. aged 20 to 34 yrs
26. any pallor
50. intent to deliver at home
3. aged 35 yrs or more
27. any jaundice or yellow eyes
51. intend to deliver at home but delivered in facility
4. was this her first pregnancy
28. any cyanosis or blue lips
52. any acute abdominal pain before labour
5. has she had 2 to 4 pregnancies
29. was baby delivered alive
53. any acute abdominal pain after delivery
6. were there >4 previous pregnancies
30. was baby delivered dead
54. any previous c-section
7. was this a multiple pregnancy
31. was baby's position abnormal
55. genital infection/foul smelling discharge pp*
8. any attempt to terminate this pregnancy
32. major bleeding in 1st 3 months of pregnancy
56. leaking membranes before labour start
9. was she <5 months pregnant at end of pregnancy
33. major bleeding >3m & before labour
57. any augmentation of labour
10. any IV or IM antibiotics required
34. major bleeding during labour
58. any persistent fever>3 wks
11. any blood transfusion required
35. major bleeding after delivery
59. any swollen glands
12. any blood transfusion received
36. was blood pressure raised during pregnancy
60. did she require iron injections
13. was she bedbound for more than 1 day pp*
37. was delivery by forceps/ventouse
61. any swelling of face
14. breathless carrying out normal activities ap*
38. was delivery by Caesarean
62. any blurred vision
15. breathless carrying out normal activities pp*
39. was delivery at home
63. any severe headache before labour
16. any loss of consciousness
40. was delivery at a health facility
64. any severe headache after delivery
17. any acute fever before pregnancy end
41. were fits only pregnancy related
65. any history of migraine
18. any acute fever after pregnancy end
42. was labour prolonged >24 hrs
66. any diagnosis of haemorrhage
19. any recurrent fever
43. was labour prolonged >48hrs
67. any diagnosis of hypertension
20. any shivering with fever
44. was delivery of the placenta delayed
68. any diagnosis of malaria
21. did she ever have fits
45. was there manual removal of the placenta
69. any diagnosis of infection
22. did she have a diagnosis of epilepsy
46. had professional assistance at delivery
70. any diagnosis of rupture
23. any hysterectomy
47. intention to deliver at health facility
71. was delivery said to be uncomplicated
24. haemoglobin less than 8g/dl
48. abnormal proteinuria reported
72. self-reported delivery complication
Hypothetical example of probabilistic interpretation of lay-reported indicators of morbidity
Probability of selected causes of death
The computer program
A computer program applying the above principles was written in Microsoft Visual FoxPro software. As illustrated in Table 2, the computer program was designed to calculate the likelihood of all-cause near-miss morbidity status separate from the likelihoods of specific causes. The concept of morbidity encompasses a spectrum of conditions that can range from very mild to life-threatening. Mild, short-term morbidity is likely to be common, but may not be of clinical significance in terms of the conditions listed in Table 1 and may be of limited public health relevance in low-income settings. Therefore, a cut-off of 30% likelihood of maternal morbidity was selected a priori, below which cases are considered to be non-morbid and the probabilities of specific causes are set to zero. The application of cut-off points to a continuous likelihood distribution is somewhat arbitrary. Nevertheless, categorical classification is useful in terms of conceptualising the severity of morbidities in clinical terms and is necessary for comparison with clinical classifications which typically uses exclusive binary categorisations of outcomes. To further enable categorical classification of cases, it is reasonable to assume that individuals whose reported symptoms result in a likelihood of near-miss in excess of 90% are at the extreme end of the morbidity spectrum whilst lower likelihoods (30% to 90%) are likely to represent clinically significant but not necessarily immediately life-threatening conditions.
For morbid cases (with a probability of near-miss above 30%), the InterSAMM computer program displays the probability of specific causes and multiple causes can be assigned to each case. Certain rules are applied to limit the number of specific causes reported. First, the likelihood of all determinate causes must have increased by an appreciable and decisive amount, defined as the square root of the unconditional cause probability. If this condition is not met for any cause due to insufficient indicators being available the cause will be categorised as 'indeterminate’. For multiple causes to be reported, each cause likelihood must fall within 50% of the previous, more likely, cause.
To handle multiple causes, each individual case can be split between multiple causes proportional to the likelihood of each determinate cause. For example, if an individual is assigned two causes, haemorrhage and anaemia, with likelihoods of 50% and 40%, respectively, 0.5 will be added to the total population count of haemorrhage, 0.4 will be added to anaemia and 0.1 (the remainder of 100% of this case) will be added to the population count of cause uncertainty. When the population counts of each cause category are divided by the total number of cases, one gets the population cause-specific morbidity fraction (CSMF) attributable to each cause category and an indication of overall uncertainty of cause diagnoses.
Development & refinement
Relevant morbidity signs and symptoms as recorded in case notes and reported by women themselves were extracted from the data and formatted to be used with InterSAMM. The population distribution of near-miss likelihoods, morbidity categorisations and cause distributions were compared with clinical classifications. Clinical diagnoses of multiple causes per case were split evenly between determinate causes, and fractions of each cause were then summed and divided by the total number of cases to calculate clinician-derived population-level CSMFs. This approach approximates the method used to handle multiple causes derived from InterSAMM, with two important differences. Firstly, cases must be split evenly between determinate causes because no quantification of likelihood of each cause is available and no assumptions of hierarchy can be assumed, even if it is likely that clinicians might consider certain causes to have a greater significance or contribution to morbidity than others. Secondly, any sense of uncertainty that clinicians had in assigning causes has been lost and cannot be accounted for.
Comparisons between clinical classifications and results from InterSAMM were carried out. An iterative process of comparisons with clinical classifications and refinements of a priori probabilities and the probabilistic model was followed, illustrated by a loop of refining the probabilistic model, re-running the data and comparing the results in Figure 1. This process enabled data-driven refinements to the a priori probabilities to produce a final probabilistic model that handled indicators to estimate morbidity and cause likelihood distributions comparable to clinical classifications.
The yellow boxes in Figure 1 illustrate the evaluation component of the study whereby additional datasets of 381 hospital deliveries from a different study in Benin (Benin II data) , and a purposive sample of 57 deliveries from a community-based cohort from Bangladesh  and 51 deliveries from a community-based cohort from India  were used to evaluate the model. The Benin II data were collected during a hospital-based validation study of an obstetric morbidity questionnaire whereby a stratified sample of women with and without maternal morbidity were identified retrospectively from case notes using criteria to define near-miss and less severe morbidity that the investigators themselves derived and described elsewhere .
The data from Bangladesh and India are population based and are a sample of women’s self-reports of their pregnancy and delivery experiences collected through interviews with mothers in their homes following the end of pregnancy as part of cluster randomised controlled trials of community-mobilisation interventions to improve maternal and neonatal outcomes. The samples from Bangladesh and India were purposefully selected to represent a range of reported morbidities and case histories, each of which was reviewed by an experienced physician who assigned likely causes of morbidity to each case.
In all evaluation data the case-mix was heavily skewed towards the morbid end of the spectrum or high-risk populations. An important difference between the evaluation datasets and the training dataset from Benin I and Burkina Faso described previously is that reports of signs and symptoms used by InterSAMM come only from women’s self-reports–hospital record data were not used as an input to the probabilistic model during the evaluation phase. InterSAMM’s performance was evaluated in terms of comparability of SAMM classifications and population CSMFs with clinical classifications. All comparisons were based on mapping and reconciliation of the range of terminologies used by clinicians to describe causes into the cause categories used by the probabilistic method. Causes identified by clinicians that did not fit into any InterSAMM cause category were grouped together as “other causes”.
Development & refinement
Population severe acute maternal morbidity (SAMM) cause distributions according to clinician classifications and probabilistic InterSAMM interpretation of data from 1734 deliveries in Benin and Burkina Faso
SAMM* > 90%
SAMM* > 30%
Non-near-miss morbid cases+
Mean (min, max) absolute difference in determinate causes compared to clinician diagnoses
Of the 1123 cases identified as uncomplicated deliveries by physicians, 25% were identified as morbid non-SAMM cases and 6% as SAMM cases by InterSAMM (Table 3). The majority of causes in these discrepant cases were puerperal infections (24%), with the remainder distributed between pre-eclampsia (14%), anaemia (12%), indeterminate causes (5%), malaria (4%) and obstructed labour (4%). The proportion of uncertainty was 38% for these cases.
Data were available for 49 of the InterSAMM indicators in the evaluation Benin dataset. The average number of indicators per case was 10 (minimum 3; maximum 22). Data from Bangladesh and India provided information for 34 and 32 indicators, respectively. The average number of indicators was 8 (minimum 4; maximum 16) in the Bangladesh data and 8 (minimum 5; maximum 13) in the India data.
Distribution and agreement between the InterVA- and clinician-derived morbidity status categories for 381 deliveries from Benin, 57 deliveries from Bangladesh and 51 deliveries from India
InterSAMM morbidity status categorisation
Frequency of reported indicators among 126 deliveries in Benin classified as uncomplicated by clinicians
Frequency of positive answers (% among clinician-assigned 'uncomplicated’ cases)
Acute fever ante-partum
Acute fever post-partum
Fever with shivering
Diagnosis of anaemia
Baby’s position abnormal
Major bleeding in early pregnancy
Major bleeding in late pregnancy
Major bleeding during labour
Major bleeding after delivery
Blood pressure raised during pregnancy
Prolonged labour >24 hours
Prolonged labour >48 hours
Delayed delivery of placenta
Manual removal of the placenta
Referral from one health centre to another
Smelly vaginal discharge
Diagnosis of hypertension
Diagnosis of infection
Self-reported delivery complication
NUMBER OF MORBIDITY INDICATORS
Population obstetric morbidity cause distributions of diagnoses by clinicians and probabilistic interpretation of data from 381 deliveries in Benin, 57 in Bangladesh and 51 in India
Mean (min, max) absolute difference in determinate causes compared to clinician diagnoses
3.4% (0.5%, 10.9%)
6.8% (1.0%, 13.7%)
4.8% (0.2%, 13.2%)
3.7% (0.3%, 10.6%)
The preliminary development and evaluation of the InterSAMM probabilistic method to estimate the burden and causes of SAMM from community-based surveys of women’s health and pregnancy experiences has highlighted potential strengths and important weaknesses of the approach. Through development and refinement of the method using training data from two settings, it was possible to produce a model that yielded population-level SAMM distributions similar to clinical assessment of the same data. When applied to different test datasets, the probabilistic method compared considerably less well with clinical classification.
Comparisons between probability-derived classifications with quantified uncertainty, and clinician classifications, which only provide absolute positive or negative diagnoses and no measure of uncertainty, is not straightforward. A probabilistic diagnosis of, say, postpartum haemorrhage with 60% likelihood (or 40% uncertainty) is not directly comparable to a physician diagnosis of the same cause with no measureable sense of certainty. Whilst clinical classifications are focused on individual cases and any uncertainty is lost when a final diagnosis is given, the quantified uncertainty of any probabilistic diagnoses can be carried over into the analysis, where the ultimate goal is to estimate population-level burdens of ill health. These differences between the two approaches to interpretation must be kept in mind. Rather than seeking to replicate the exact distributions of SAMM cases and causes derived by clinicians, evaluation focussed on achieving plausible distributions of SAMM and non-SAMM cases and cause distributions that adequately represent burdens of morbidity at the population-level and would be equally as valuable in guiding policy or intervention decisions.
Discrepancies between InterSAMM and clinical classifications highlight potential weaknesses of the current model, for which further thought and revision is needed. For example, when using the test data from Benin, InterSAMM identified considerably more morbid cases than clinicians. This is the crucial common problem with the analysis of self-reported maternal morbidity data and obviously the preliminary InterSAMM method has not overcome problems of low specificity. However, scrutiny of discrepant cases, in which the probabilistic method identified morbidity but clinicians did not, shows that the majority reported symptoms suggestive of some degree of complication. As such, the conclusions reached by the probabilistic method may not be unreasonable from a purely symptom-based probability approach. The challenge remains, however, to move beyond this purely symptom-based approach to improve specificity by further understanding the uncertainty of self-reported symptoms and building this uncertainty, as well as measures of severity, into InterSAMM.
The fact that physicians categorised certain cases as uncomplicated, despite morbidity indicators being reported (Table 5), reflects a divergence in women’s perceptions of childbirth from medical diagnoses of normal labour. This is not surprising, perhaps, but is important from the point of view of population-level measurement from community-based surveys. Clinicians in the Benin and Burkina Faso datasets used strict diagnostic protocols to interpret medical records written by other clinicians to reach a diagnosis. There may have been information available in the hospital records that did not form part of the clinician’s diagnostic criteria for complications, yet may have been used by the probabilistic method. Clinician review of the data from Bangladesh and India did not employ strict diagnostic criteria, with the coding physician being free to diagnose as many potential causes of morbidity as they deemed appropriate from the available interview data, which, in contrast to the West African setting, resulted in more cases being classified as morbid by the clinician than by the probabilistic method. The Benin and Burkina Faso clinicians are also likely to have utilised diagnostic criteria such as blood pressure measurements and clinical observations that are not part of the probabilistic method’s input indicators. Furthermore, symptoms may have been absent from hospital records, particularly if they occurred antepartum or post-discharge, and so were not available for clinical diagnoses but were available to the probabilistic method. These factors may further explain observed discrepancies and may highlight a need to further align InterSAMM with clinical criteria.
Cause-specific discrepancies, such as the varying proportions of infection in the test datasets, may relate to varying definitions of such complications and also to the fact that broad physical symptoms that may be experienced by women during childbirth or postpartum, including shivering or vaginal discharge, can affect the specificity of diagnosis. Greater understanding of the way that clinicians interpreted and valued certain indicators, and how or why clinicians and/or the diagnostic criteria they used excluded certain reported signs and symptoms as indications of morbidity, may be helpful in future refinements of InterSAMM. Similarly, insight into reporting biases in community-based maternal morbidity surveys may inform future refinements. Such information is likely to be useful in establishing a priori probabilities for indicators which are frequently over-or mis-reported and should therefore have a lesser influence on raising the likelihood of complications and their causes.
The fact that absolute differences between CSMFs for specific causes varied between settings highlights the problem of variability of diagnoses between settings and coding clinicians as, given the completely standardised way in which the automated probabilistic method handles symptom data, one would expect absolute differences to be fairly consistent if all other factors were held constant. A degree of inter-and intra-rater variability in diagnosing morbidities is inevitable–indeed it is part of the motivation for the development of a standardised method–but it means that comparisons of new methods against inconsistent and potentially flawed reference standards (in which the absolute “gold-standard” diagnoses are difficult to obtain) must be interpreted with caution. Variability in the tools used to collect symptom data from women may have further limited the comparability of results from different settings, as may varying degrees of recall and reporting bias whereby respondents’ answers may be influenced by recollection of events and the perceived desirability of answers . Whilst the probabilistic approach used by InterSAMM may be better able to handle uncertainty in reported indicators due to recall or reporting bias than non-probabilistic methods, any future developments of InterSAMM may need to consider the effect of differing data capture processes and questionnaires and the effect of differing availability of indicators.
Previous work on verbal autopsies has shown that the probabilistic InterVA method for cause of death ascertainment is relatively insensitive to minor variations in the prior probabilities . The same is likely to be true for the probabilistic method applied to maternal morbidity in this study, and may explain why “ball-park” probabilities in the current model and refinements using the hospital-based, training datasets from Benin and Burkina Faso, were sufficient to create a workable model to explore the method’s potential utility. Nevertheless, should the method be developed further, a more sound approach to establishing a priori probabilities should be used. Work on verbal autopsies has successfully used a system of expert consensus to approximate underlying probabilities [12, 17, 18], and, given the lack of existing, reliable data on the burden and causes of obstetric morbidity in communities where many women deliver at home, a similar approach may be appropriate for InterSAMM. Further understanding of how women perceive and describe morbidity symptoms and delivery complications could also benefit the development of the probability matrix and the indicators used, perhaps involving input from women themselves or birth attendants. Finally, there may be a need for contextual variations in a priori probabilities where, rather than a protective factor, for example, delivery at a facility may indicate a complication in a population that normally delivers at home.
None of the data used during the development and preliminary evaluation of InterSAMM are representative of a general population of recently-delivered women, in which one would expect the vast majority to have had uncomplicated pregnancies, deliveries and post-partum periods. InterSAMM’s performance in a population for which it is intended remains untested. Further testing on data from a range of settings is important, although sourcing suitable datasets with adequate reference diagnoses for comparison has proved challenging. More detailed explorations of diagnostic accuracy at the individual and population level, such as sensitivity, specificity and positive predictive values, or assessments of inter-rater agreement, may be appropriate in future evaluations, but, once again, their interpretation must be grounded in the realities of the reference standards being used. Newly proposed chance-corrected concordance and diagnostic accuracy measures that take into consideration the potential for random agreement between methods that can generate multiple causes from a finite list of causes may also be useful in future evaluations . All evaluations, however, must relate to the intended use of InterSAMM to estimate population levels of morbidity and its causes, whereby shortcomings in accuracy may be offset by plausibility, efficiency, adequacy for purpose and advantages of unique reliability. There may also be a need for a compromise between strict diagnostic criteria, as would be used in clinical settings, and broader conceptualisation of morbidity as deemed important by women themselves and important to population-level understandings of SAMM to inform public health.
The preliminary probabilistic InterSAMM method described and evaluated here has important limitations, but shows promise in overcoming longstanding barriers to standardised measurement of maternal morbidity that allows for the uncertainty of women’s self-reported symptoms in retrospective interviews. Further development, refinement and evaluation, as well as exploration of other statistical methods [20–22], is likely to be worthwhile for its potential to advance the measurement of maternal morbidity and revealing population burdens and causes of severe acute maternal morbidity.
aAggregated broad cause categories were as follows: Pre-eclampsia/Eclampsia = Pregnancy Induced Hypertension (PIH); Uterine rupture/pre-rupture/Obstructed labour = Dystocia; Ante-partum/post-partum haemorrhage = Haemorrhage; Genital infection/Malaria/Other infection = Infection.
Severe Acute Maternal Morbidity
Interpreting Severe Acute Maternal Morbidity
Cause Specific Morbidity Fraction.
This work was supported by the Child Health Epidemiology Reference Group (CHERG).
EF and PB are supported by FAS, the Swedish Council for Working Life and Social Research (grant 2006-1512). DO is supported by a Wellcome Trust Senior Research Fellowship in Clinical Science (091561).
Collaboration between UCL (UK), Ekjut (India) and BADAS-PCP (Bangladesh) is supported by a Wellcome Trust Strategic Award (085417ma /Z/08/Z).
The authors would like to thank João Paulo Dias De Souza and Maria Quigley for their constructive review and useful comments on an early presentation of this work.
We are grateful to all data teams in each country and to the women who participated in each study.
- Merdad L, Hill K, Graham W: Improving the measurement of maternal mortality: the sisterhood method revisited. PLoS One. 2013, 8: e59834. 10.1371/journal.pone.0059834PubMed CentralView ArticlePubMedGoogle Scholar
- Say L, Souza JP, Pattinson RC: Maternal near miss–towards a standard tool for monitoring quality of maternal health care. Best Pract Res Clin Obstet Gynaecol. 2009, 23: 287-296. 10.1016/j.bpobgyn.2009.01.007View ArticlePubMedGoogle Scholar
- Filippi V, Ronsmans C, Campbell OMR, Graham WJ, Mills A, Borghi J, Koblinsky M, Osrin D: Maternal health in poor countries: the broader context and call for action. Lancet. 2006, 368: 1535-1541. 10.1016/S0140-6736(06)69384-7View ArticlePubMedGoogle Scholar
- Hardee K, Gay J, Blanc AK: Maternal morbidity: neglected dimension of safe motherhood in the developing world. Global Public Health. 2012, 7: 603-617. 10.1080/17441692.2012.668919PubMed CentralView ArticlePubMedGoogle Scholar
- Filippi V, Goufodji S, Sismanidis C, Kanhonou LG, Fottrell E, Ronsmans C, Eusebe A, Patel V: Effects of severe obstetric complications on women’s health and infant mortality in Benin. Trop Med Int Health. 2010, 15: 733-742. 10.1111/j.1365-3156.2010.02534.xPubMed CentralView ArticlePubMedGoogle Scholar
- Filippi V, Ganaba R, Baggaley RF, Marshall T, Storeng KT, Sombié I, Ouattara F, Ouedraogo T, Akoum M, Meda N: Health of women after severe obstetric complications in Burkina Faso: a longitudinal study. Lancet. 2007, 370: 1329-1337. 10.1016/S0140-6736(07)61574-8View ArticlePubMedGoogle Scholar
- Ronsmans C, Scott S, Adisasmita A, Deviany P, Nandiaty F: Estimation of population-based incidence of pregnancy-related illness and mortality (PRIAM) in two districts in West Java, Indonesia. Br J Obstet Gynaecol. 2009, 116: 82-90. 10.1111/j.1471-0528.2008.01913.x. 10.1111/j.1471-0528.2008.01913.xView ArticleGoogle Scholar
- Filippi V, Ronsmans C, Gandaho T, Graham W, Alihonou E, Santos P: Women’s reports of severe (near-miss) obstetric complications in Benin. Stud Fam Plan. 2000, 31: 309-324. 10.1111/j.1728-4465.2000.00309View ArticleGoogle Scholar
- Ronsmans C, Achadi E, Cohen S, Zazri A: Women’s recall of obstetric complications in South Kalimantan, Indonesia. Stud Fam Plan. 1997, 28: 203-214. 10.2307/2137888. 10.2307/2137888View ArticleGoogle Scholar
- Stewart M, Festin M: Validation of women’s reporting and general recall of major obstetric complications treated at the Philippines General Hospital. Int J Obstet Gynaecol. 1995, 48: S53-S66.View ArticleGoogle Scholar
- Montgomery A, Goufodji S, Kanhonou L, Alihonou E, Collin S, Filippi V: Validity and reliability of postpartum morbidity questionnaires in Benin. Matern Child Health J. 2012, 16: 1728-1731. 10.1007/s10995-011-0859-9View ArticlePubMedGoogle Scholar
- Fottrell E, Byass P, Ouedraogo TW, Tamini C, Gbangou A, Sombie I, Hogberg U, Whitten KH, Bhattacharya S, Desta T, Deganus S, Toruni J, Fitzmaurice AE, Meda N, Graham WJ: Revealing the burden of maternal mortality: a probabilistic model for determining pregnancy-related causes of death from verbal autopsies. Popul Health Metrics. 2007, 5 (1).
- Fottrell E, Kanhonou L, Beghague DP, Patel V, Filippi V: Risk of psychological distress following severe obstetric complications in Benin: the role of economics, physical health and spousal abuse. Br J Psychiatry. 2010, 196: 18-25. 10.1192/bjp.bp.108.062489PubMed CentralView ArticlePubMedGoogle Scholar
- Azad K, Barnett S, Banerjee B, Shaha S, Khan K, Rego AR, Barua S, Flatman D, Pagel C, Prost A, Ellis M, Costello A: Effect of scaling up women’s groups on birth outcomes in three rural districts in Bangladesh: a cluster-randomised controlled trial. Lancet. 2010, 375: 1193-1202. 10.1016/S0140-6736(10)60142-0View ArticlePubMedGoogle Scholar
- Tripathy P, Nair N, Barnett S, Mahapatra R, Borghi J, Rath S, Rath S, Gope R, Mahto D, Sinha R, Lakshminarayana R, Patel V, Pagel C, Prost A, Costello A: Effect of a participatory intervention with women’s groups on birth outcomes and maternal depression in Jharkhand and Orissa, India: a cluster-randomised controlled trial. Lancet. 2010, 375: 1182-1192. 10.1016/S0140-6736(09)62042-0View ArticlePubMedGoogle Scholar
- Fottrell E, Kahn K, Tollman SM, Byass P: Probabilistic methods for verbal autopsy interpretation: InterVA robustness in relation to variations in a priori probabilities. PLoS One. 2011, In Press.Google Scholar
- Byass P, Fottrell E, Huong DL, Berhane Y, Corrah PT, Kahn K, Muhe L, Van DD: Refining a probabilistic model for interpreting verbal autopsy data. Scand J Public Health. 2006, 34: 26-31. 10.1080/14034940510032202PubMed CentralView ArticlePubMedGoogle Scholar
- Byass P, Chandramohan D, Clark S, D’Ambruoso L, Fottrell E, Graham W, Herbst A, Hodgson A, Hounton S, Kahn K, Krishnan A, Leitao J, Odhiambo F, Sankoh O, Tollman S: Strengthening standardised interpretation of verbal autopsy data: the new InterVA-4 tool. Global Health Action. 2012, 5: 19281.Google Scholar
- Murray CJL, Lozano R, Flaxman AD, Vahdatpour A, Lopez AD: Robust metrics for assessing the performance of different verbal autopsy cause assignment methods in validation studies. Popul Health Metrics. 2011, 9: 28-10.1186/1478-7954-9-28View ArticleGoogle Scholar
- Murray CJL, James SL, Birnbaum JK, Freeman MK, Lozano R, Lopez AD: Simplified symptom pattern method for verbal autopsy analysis: multisite validation study using clinical diagnostic gold standards. Popul Health Metrics. 2011, 9: 30-10.1186/1478-7954-9-30View ArticleGoogle Scholar
- James SL, Flaxman AD, Murray CJL: Performance of the tariff method: validation of a simple additive algorithm for analysis of verbal autopsies. Popul Health Metrics. 2011, 9: 31-10.1186/1478-7954-9-31View ArticleGoogle Scholar
- Flaxman AD, Vahdatpour A, Green S, James SL, Murray CJL: Random forests for verbal autopsy analysis: multisite validation study using clinical diagnostic gold standards. Popul Health Metrics. 2011, 9: 29-10.1186/1478-7954-9-29. 10.1186/1478-7954-9-29View ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.