Risk factors for obstructed labour in Eastern Uganda: A case control study

Introduction Obstructed labour (OL) is an important clinical and public health problem because of the associated maternal and perinatal morbidity and mortality. Risk factors for OL and its associated obstetric squeal are usually context specific. No epidemiological study has documented the risk factors for OL in Eastern Uganda. This study was conducted to identify the risk factors for OL in Mbale Hospital. Objective To identify the risk factors for OL in Mbale Regional Referral and Teaching Hospital, Eastern Uganda. Methods We conducted a case control study with 270 cases of women with OL and 270 controls of women without OL. We consecutively enrolled eligible cases between July 2018 and February 2019. For each case, we randomly selected one eligible control admitted in the same 24-hour period. Data was collected using face-to-face interviews and a review of patient notes. Logistic regression was used to identify the risk factors for OL. Results The risk factors for OL were, being a referral from a lower health facility (AOR 6.80, 95% CI: 4.20–11.00), prime parity (AOR 2.15 95% CI: 1.26–3.66) and use of herbal medicines in active labour (AOR 2.72 95% CI: 1.49–4.96). Married participants (AOR 0.59 95% CI: 0.35–0.97) with a delivery plan (AOR 0.56 95% CI: 0.35–0.90) and educated partners (AOR 0.57 95% CI: 0.33–0.98) were less likely to have OL. In the adjusted analysis, there was no association between four or more ANC visits and OL, adjusted odds ratio [(AOR) 0.96 95% CI: 0.57–1.63)]. Conclusions Prime parity, use of herbal medicines in labour and being a referral from a lower health facility were identified as risk factors. Being married with a delivery plan and an educated partner were protective of OL. Increased frequency of ANC attendance was not protective against obstructed labour. We conducted this study in the labour suite of Mbale regional referral Hospital in Eastern Uganda. This hospital, serves 14 districts in the Elgon zone with an estimated population of 4 million people. This is a government run, not-for-profit, charge-free, 470-bed hospital with 52 maternity beds. Annually, about 12,000 childbirths occur in this hospital with a caesarean section rate of 35% and nearly 500 mothers have OL. About two thirds of these mothers with OL are referrals in active labour from the lower health units. Unmatched case control design with incidence density sampling of the controls admitted in the same delivery suite. All patients admitted to the labour suite in active labour at term (≥ 37 weeks of gestation) were screened. A Medical Officer or Obstetrician diagnosed OL using the American College of Obstetricians and Gynecologists (ACOG) guideline for arrest of labour [14] and local protocols. A case was defined as; a cervical dilatation ≥ 6cm with ruptured membranes, having adequate contractions lasting > 4hrs with no change in cervical dilatation in the first stage of labour. For the second active stage of labour, arrest was defined as a delay of > 2 hours for the nullipara and > 1 hour for the multipara with adequate uterine contractions. In addition, a case had to have any two of the following obvious signs of severe obstruction: caput formation, Bandl’s ring, sub-conjunctival hemorrhages and edematous vulva. Controls were women admitted to the labour suit within the same 24-hour period in active labour without obstruction. We used the formula described by Fleiss with a continuity correction to estimate the sample size[15]. The exposure factor was the proportion of pregnant women who attended < 4 ANC visits. We enrolled 270 cases and 270 controls based on the following assumptions: two-sided 95% confidence level, power of 95%, ratio 1:1 to detect an odds ratio of at least 2 for the risk of OL among pregnant women who attended < 4 ANC visits as the main exposure variable[16–18]. We further assumed that controls were like any other pregnant woman in Uganda who attended at least 4 ANC visits (60%) according to the Uganda demographic and health survey [9]. We consecutively enrolled all eligible incident cases between July 2018 and February 2019. We used simple random sampling to select one control from a list of admissions in active labour immediately after enrolling each case. Before recruitment, all respondents gave us written informed consent and pregnant adolescents below the legal age of 18 years were taken as emancipated minors[19]. We used unique study numbers issued at enrolment to identify each respondent. Cases were women with OL carrying singleton, term pregnancies in cephalic presentation. Controls were women in active labour without obstruction carrying singleton, term pregnancies in cephalic presentation. We excluded women with other obstetric emergencies such as antepartum haemorrhage, Pre-eclampsia and eclampsia (defined as elevated blood pressure of at least 140/90 mmHg, urine protein of at least 2+, any of the danger signs and fits), premature rupture of membranes and intrauterine fetal death. We also excluded all women from outside the Hospital catchment area of 14 districts as either cases or controls. The socio-demographic factors highlighted in the literature to predispose women to OL were the participant’s age, marital status, occupation, level of education, the occupation and education level of the spouse as well as distance to the nearest health facility and the place of residence[10,12,17,20,21]. The obstetric factors were gravidity, number of ANC visits, having a delivery plan in place, a history of being referred from a lower health facility and use of herbal medications during labour[16,17]. Physical examination included the respondent’s height and fetal birth weight. Our main exposure was the number of ANC visits attended as indicated on the ANC card, the other covariates were considered as confounders. We used an interviewer-administered questionnaire running on an open data kit (ODK) platform. Trained research assistants (RA’s) who are qualified midwives administered the questionnaire to all participants in the local dialect. We blinded all the RA’s to the hypothesis of the study. Available records such as the antenatal cards, facility registers and case report files were reviewed by the RA’s to crosscheck some of the verbal responses. The principal investigator (PI) would, on a daily basis access and review the data from the Google Aggregate server for completeness. The data was uploaded to a password protected server to which only the PI or his designee had access. Assisted by a statistician, the data was downloaded into an excel spreadsheet and exported to Stata version 14 for further cleaning and analysis. Baseline socio-demographic, physical and obstetric characteristics of the cases and controls were compared, to identify any differences. Normality of the continuous variables was tested for using the Shapiro-Wilk test. We summarised continuous variables using means and standard deviations. Whereas frequencies and percentages were used for the categorical variables. We used logistic regression (LR) to estimate Odds ratios, and 95% confidence intervals to examine the association between the number of ANC visits (< 4 Vs ≥ 4) and the different socio-demographic, physical and obstetric covariates in bivariable and multivariable analysis. We included all factors that are known to confound the relationship between the frequency of ANC attendance and OL in the multivariable LR model, based on biological plausibility. In order to control for potential residual confounding due to factors that we had not previously hypothesized to be confounders, we also included those variables for which bivariable analysis returned a p-value equal to or less than 0.25. We reasoned that a cut-off of 0.25 would allow us to test the effect of any factors previously not known to have a confounding effect on the relationship between OL and the frequency of ANC attendance, without including those factors that were reasonably least likely [22]. Multicollinearity between explanatory variables was assessed using the variance inflation factor (VIFs), and they were all less than 1.5. In the final adjusted multivariable model, we included all the statistically significant covariates (being a referral, a history of using herbal medicines, having a delivery plan, prime parity and partner education level). Confounding was considered present, if the difference between the crude and adjusted OR was ≥ 10 percentage points[23,24]. The Makerere University School of Medicine Research and Ethics Committee (#REC REF 2017–103) and the Uganda National Council for Science and Technology (HS217ES) approved the protocol. The Mbale Hospital Research and Ethics Committee (MRRH-REC IN-COM 00/2018) gave us administrative clearance. The hospital protocols were followed in management emergencies during the study.