Dr A. Dumont, Hôpital Sainte-Justine, 3175 Côte Sainte-Catherine, H3T-1C5 Montréal, Québec, Canada.
Objective To evaluate, with volunteer professionals in a resource-poor setting, an approach of audit and feedback to promote local implementation of emergency obstetric guidelines.
Design Triple cohort observational time series study.
Setting A 46-bed obstetric unit in an academic-affiliated community hospital in Senegal.
Population All pregnant women with haemorrhagic and hypertensive complications who were admitted to the maternity unit during the study periods.
Methods To assess the benefits of guidelines implementation, maternal outcomes during the intervention period were compared with those occurring in two one-year periods when staff daily supervision was the main potentially effective action on clinical management.
Main outcome measures The intervention strategy was criteria-based audits with regular feedback over a one-year period. The clinical focus was haemorrhage and hypertension the most frequent causes of maternal death in the study population. Hospital charts were audited by external reviewers. The primary outcome was the case fatality rate (CFR) among patients with haemorrhage and hypertension.
Results There was an increase in morbidity diagnoses during the intervention period. In addition, there was a marked increase in obstetric interventions, especially for transfusions and caesarean deliveries. Patients characteristic-adjusted case fatality decreased by 53% between baselines I and II and during the intervention period by 33% and 24%, compared with baseline periods I and II, respectively. Outcome improvements were different for haemorrhage and hypertension.
Conclusion While staff daily supervision may have improved maternal outcome before the intervention period, audit and feedback produced marked effects on emergency obstetric care, specially for complications requiring highly trained management (e.g. pre-eclampsia). Audit and feedback are one of the potentially effective guidelines implementation strategies that should be considered for further studies in resource-poor health facilities.
Maternal mortality is a major public health problem, especially in West Africa where maternal mortality ratios are still very high.1 Most maternal deaths occur during—or few hours after—delivery. Haemorrhage, hypertension, obstructed labour and sepsis are the major direct obstetric causes.2 The treatments for those obsterical complications are well known and appropriate emergency obstetric care should prevent most of these deaths.3 Situation analyses in various heath facilities in West Africa underlined problems related to health services management and to staff attitudes.4 The main factors identified as being responsible for poor quality of care were failure to offer 24-hour services, lack of drugs and supplies and low competence of birth attendants. Inadequate preservice and in-service training, lack of technical support and supervision and absence of standard treatment guidelines could explain the heterogeneous quality of care provided by midwifes in first line and referral hospitals.5 Furthermore, anthropological studies in West Africa pointed out that while both pregnant women and midwives are influenced by the same social rules (e.g. linguistic taboos, respect and shame), technical constraints force midwives to violate those rules, making the application of their technical skills very difficult.6 Thus, midwives and other staff involved in delivery management must learn how to implement modern obstetric guidelines within specific cultural environments.
The movement to develop and disseminate clinical practice guidelines (CPGs) has been well established in industrialised countries for more than a decade.7 Practice guidelines, however, although useful as one component in quality assurance programme, do not by themselves appear to change professional behaviour, which has proven resistance to outside directives.8 Supplementary local activities are needed for implementing CPGs. In resource-high settings, audits and feedback, opinion leaders, reminder systems and academic detailing were shown to be moderately or strongly effective.9 In developing countries, the audit approach is a promising strategy that could be applied at the local level to change obstetric practice.10–14 Few studies have shown the impact of this complex intervention on maternal mortality.
This ‘before-and-after’ study evaluates, with community hospital midwives and physicians, criteria-based audits with feedback for encouraging local implementation of emergency obstetric guidelines. The clinical focus was haemorrhage and hypertension the most frequent causes of maternal death in the study population. The objective was to evaluate whether this strategy would, under usual clinical circumstances with volunteer professionals, lead to changes in obstetric practices and to a decrease in maternal mortality.
We conducted an observational study with two baseline periods and an intervention period. The study was approved by the local committee on clinical research in the Centre de Santé Roi Baudouin, Guédiawaye, Dakar, Senegal. The intervention had a one-year duration (1 March 2001 through 28 February 2002). Two baseline periods of equal duration were used for comparison of clinical outcomes. Baseline period I was selected as an appropriate time interval after the opening of the surgical unit (1 January 1998, to 31 December 1998), and baseline period II was selected to evaluate possible time-related changes in outcomes before the study intervention (1 January 1999, to 31 December 1999). Year 2000 was the preintervention period for the elaboration and dissemination of CPGs and thus was neither baseline period nor an intervention period.
All pregnant women with haemorrhagic and hypertensive complications who were admitted to the maternity unit during the study period were included. The following complications were examined: placental abruption, haemorrhagic placenta previa, postpartum haemorrhage, pre-eclampsia and eclampsia. Women who were referred from other health facilities were included in the study unless they met predefined inclusion criteria. The patients with haemorrhage or hypertensive disorders who were transferred from the study site to other hospitals were included in the study and information was obtained by telephone within six weeks after the transfer. We excluded women with first trimester complications (unsafe abortion, septic or haemorrhagic miscarriage, ectopic pregnancy) because the CPGs focussed on second and third trimester complications and women admitted after 24 hours following delivery because care of unit was not concerned at the beginning.
The study site was a surgical maternity unit in a 72-bed academic-affiliated district hospital located in the suburb of Dakar, Senegal's capital. The surgical unit was built in 1997. Three residents, four nurse–anesthesists and one gynaecologist–obstetrician provided 24-hour surgical obstetric care. Labour management and normal deliveries were attended by 19 midwifes. At the daily meeting with staff, the senior gynaecologist–obstetrician reviewed all patients' charts to control for clinical data and assist with triage issues. During the study periods, he identified obstetric complications and completed missing information, if necessary.
Twelve primary care facilities usually referred patients (∼15%) to this centre. The 46-bed obstetric unit provides care for low and high risk pregnancy but not intensive care patients who need artificial ventilation, dialysis or other invasive intensive care. If necessary, mothers (∼1%) or babies (∼5%) were referred to the university hospital located 15 km from the study centre.
The CPGs addressed the five selected complications: placental abruption, haemorrhagic placenta previa, postpartum haemorrhage, pre-eclampsia and eclampsia. Multiple sources of information were used to develop internal CPGs: (i) guides for the ‘management of pregnancy and delivery complications’ recommended by the World Health Organisation (http://www.who.int/reproductive-health/); (ii) national guidelines (‘Politiques, Normes et protocoles en matière de santé de la reproduction’, Ministère de la Santé, Dakar 2000); and (iii) relevant international literature. A first draft was submitted to midwifes and physicians for review and suggestions. The final versions of CPGs were mailed to each professional and handbooks were available in each care unit. In addition, explicit criteria were defined for the management of each complication. Explicit criteria did not exceed 12 items. For example,Fig. 1 shows the 11 criteria to evaluate antepartum eclampsia management.
During the intervention period, in addition with daily staff supervision, a team of external reviewers (two midwives and one physician) conducted a clinical audit of hospital charts for all women with haemorrhagic and hypertensive complications. Clinical data acquisition and charts audit were two separate processes to avoid interference with staff supervision by the senior gynaecologist–obstetrician. Audit and feedback concerned the following activities: (1) each week, all selected charts were audited by external reviewers who compared actual practice with the explicit criteria; (2) every three months, the entire department (midwives and physicians) held meetings for feedback and discussion of the audit results, especially discrepancies between CPGs and actual practice.
Feedback was given by one of the authors (AD). The same study team member presented the feedback every three months as well as information on any subsequent discussion of discrepancies between CPGs and practice. The process was facilitated by the department chair. The main feedback elements were diagrams showing the proportions of standard care by complications (e.g. placental abruption, haemorrhagic placenta previa, postpartum haemorrhage, pre-eclampsia and eclampsia) and according to the specific step in the clinical management (e.g. initial assessment, diagnosis, treatment and monitoring). Feedback was at the department level and no individual feedback was proposed in order to respect confidentiality.
A member of the study team (AD) extracted the clinical data from computerised patients' charts. All patients' charts were controlled by a senior obstetrician and computerised after patients release by a trained administrative assistant using Epi Info 6.01 software (Centers for Disease Control, Atlanta, Georgia, USA). A descriptive analysis was performed on maternal characteristics and practice patterns. Maternal characteristics included conditions known to be associated with poor maternal outcome to control for possible changes in this hospital-based population during study periods: age, parity, previous caesarean section, number of prenatal visits and referrals from other health facilities. The profile of medical practice pertains to the following interventions: labor induction using oxytocin or intravaginal prostaglandins; intrapartum oxytocin use; transfusion (packed red blood cells or fresh frozen plasma); and caesarean section. We compared maternal characteristics and medical practice between study periods using χ2 tests for dichotomous variables and linear regression for continuous variables.
The primary outcome measure was the case fatality rate (CFR) among patients with haemorrhagic and hypertensive complications. The CFR is the proportion of women deceased from a specific complication (primary cause of death) among the women with the corresponding complication. This indicator is expected to reflect the level of quality of care. The primary cause of death was determined by two senior obstetricians who reviewed all maternal death occurring in the study site. Differences between CFR across time periods were measured using crude odds ratios (ORs) with 95% confidence intervals. Then we calculated corresponding adjusted ORs to control for possible changes in maternal characteristics and medical practice.
A total of 446 patients from baseline period I, 520 from baseline period II and 712 from intervention period had haemorrhagic or hypertensive complications and met the inclusion criteria's. Patients were excluded because of unsafe abortion, septic or haemorrhagic miscarriages (n= 154), ectopic pregnancies (n= 77) and admission after 24 hours following delivery (n= 10). Patient characteristics are shown in Table 1. Groups were similar with respect to age and parity but differed for all other variables: previous caesarean section, antenatal care and referral for delivery. The increase in diagnosed morbidity during the intervention period was mainly explained by the increase in pre-eclampsia diagnoses (274 during the intervention period vs 83 and 140 in baseline periods I and II, respectively).
Table 1. Demographic and obstetric characteristics of the patient population. Values are presented as mean [SD] or n (%).
The total of obstetric complications is different from the total of women with complications because patients could present two diagnoses and more simultaneously.
3 and more
Previous caesarean section
3 and more
Referral for delivery
Data on medical practice are shown in Table 2. Significant differences between time periods were observed for transfusion and caesarean section rates. There were no significant changes for labour inductions, oxytocin use and forceps.
Table 2. Medical practice. Values are given as number of interventions (%) in the patient population.
Proportions of substandard care during the intervention period varied by complication (results not shown): postpartum haemorrhage (63%), pre-eclampsia (52%), eclampsia (46%), placental abruption (42%) and placenta previa (33%); and according to the specific step in clinical management: initial assessment (23%), diagnosis (23%), treatment (46%) and monitoring (75%).
The total of registered maternal deaths for the three periods were 47, 37 and 29 and the number of women deceased of haemorrhage and hypertension was 32, 23 and 20, respectively. Changes in CFRs are shown in Table 3 and Fig. 2. Comparison of crude ORs demonstrates a decrease in CFRs for haemorrhage and hypertension between baselines I and II and during the intervention period compared with baselines I and II, respectively.
Table 3. Maternal outcome. Values are given as crude and adjusted* OR (95% CI) measuring the differences between CFRs for the study periods.
OR adjusted for age, parity, previous caesarean section, number of prenatal visits, referral for delivery, transfusion and caesarean delivery.
Case fatality for haemorrhage
Case fatality for hypertension
Case fatality for haemorrhage and hypertension
Outcome improvements were different for haemorrhage and hypertension. For haemorrhagic complications only, the decrease in CFRs was particularly marked between baseline period I and baseline period II. For hypertensive disorders only, the decrease was particularly marked during the intervention period. When we controlled for maternal characteristics and medical practice changes (transfusions and caesarean sections), adjusted ORs showed similar trends in CFRs across time periods. After multivariate analysis, CFR for haemorrhagic and hypertensive complications decreased significantly by 53% between baselines I and II, and during the intervention period by 33% (significant) and 24% (non-significant), compared with baseline periods I and II, respectively.
This study assessed the effectiveness of one promising behaviour change strategy in encouraging implementation of CPGs for emergency obstetrics in developing countries. When applied to 27 qualified professionals, especially to physicians and midwives in a community hospital in Senegal, the criteria-based audit and feedback for haemorrhage and hypertensive complications, associated with staff daily supervision, produced marked effects on practice patterns and maternal outcomes.
The increase in diagnosed morbidity may result from the CPGs implementation. Indeed, the guidelines have focussed not only on the management of complications, but also on their diagnosis. The effects may be greater for hypertensive disorders (pre-eclampsia) than for haemorrhagic complications. Haemorrhage is a diagnosis easily made by professionals, while diagnosing pre-eclampsia requires more highly trained observation of the patient (e.g. patient history, signs of high blood pressure and urine albumin measurement). These findings, which are consistent with those of other authors,4,15,16 suggest that trained staff detect more obstetric complications than do non-trained professionals.
A second effect of the audit and feedback was the increase in medical interventions. Among the study population, the main indications for transfusions (e.g. placental abruption, placenta previa and postpartum haemorrhage) and for caesarean deliveries (e.g. placental abruption, placenta previa, pre-eclampsia and eclampsia) were precisely the ones the guidelines have focussed on. Several factors should be taken into consideration in the interpretation of these results. First, baseline transfusion and caesarean section rates should be suboptimal before the intervention period. Secondly, the audit and feedback helped to identify processes of care that should be considered a priority. The audit revealed that the main factor responsible for non-compliance was the lack of monitoring. Improving monitoring during the intervention period may have forced professionals to perform more caesarean sections and transfusions to improve maternal outcome.
However, audit and feedback strategy to implement CPGs in the present study could not totally explain the improvement in maternal outcome because patient characteristics-adjusted case fatality started to decrease within the two baseline periods. These results suggest that other interventions, which were implemented during baseline period I or II, had an impact on clinical management. During these two periods, the main potentially effective intervention on clinical management was the daily staff supervision by a senior gynaecologist–obstetrician. Every morning, the senior was expected to review all patients' charts. Obstetric complications were identified and missing information completed, if necessary. In addition, the senior may define management plans and assist with triage issues. It is possible that this daily supervision may have modified the clinical management plans before the intervention period. The impact on maternal outcome was different for haemorrhage and hypertension because hypertensive disorders were less detected during the two baselines periods, compared with the intervention period, while haemorrhage diagnoses did not significantly change across time periods. Thus, audit and feedback, when associated with daily supervision, probably helped professionals to better detect severe hypertension and to appropriately manage the complication.
Jamtved et al.17 reviewed the effects of audit and feedback on professional practice and health care outcomes. The authors included 85 studies from various health services, and most of these were done in industrialised countries. Non-compliance rates with desired practice varied considerably from a 9% increase in non-compliance to 71% decrease in non-compliance. Baseline non-compliance was a factor that appeared to predict the effectiveness across studies. The absolute effects of audit and feedback are more likely to be large when baseline adherence to recommend practice is low. Only one randomised controlled trial of audit and feedback has been carried out in obstetric services. Lomas et al.18 demonstrated that opinion leader was more effective to reduce caesarean section rate among women with previous caesarean section, compared with audit and feedback in 16 community hospitals in Canada. Few trials have investigated the effect of the audit and feedback process on obstetric care in developing countries. Wagaarachchi et al.11 have shown that criterion-based clinical audit in four district hospitals in Ghana and Jamaica is a feasible and acceptable method for quality assurance. However, the authors did not show the impact of this strategy on maternal outcome.
These findings shows that the guidelines implementation in resource-poor settings is feasible. While staff daily supervision may have improved maternal outcome before the intervention period, audit and feedback produced marked effects on emergency obstetric care, specially for complications requiring highly trained management (e.g. pre-eclampsia). Audit and feedback is one of the potentially effective implementation strategies that should be considered for further studies in such settings. Further studies are needed to understand more precisely the forces and variables influencing emergency obstetric practice in developing countries and to confirm the effectiveness of the audit approach to change staff behaviour.
This study was supported by the French Cooperation Office, Ambassade de France au Sénégal. The authors would like to thank William Fraser and Joane Delage for reviewing this article and the midwives and doctors in the Roi Baudouin community hospital for their cooperation.