Study setting and design
Kigoma Region, located in the northwest corner of Tanzania by Lake Tanganika, covers 45,066 km2 and had a population of 2,127,930 in 2012. Approximately 83% of the population is classified as rural with farming as the primary economic activity [42]. In 2015, emergency obstetric and neonatal care facilities provided care for 83% of all direct obstetric complications. Eight out of ten maternal deaths in facilities were due to direct obstetric causes in 2011–2015 [43].
A regionally representative multistage survey of reproductive age women (15–49 years) was conducted in July–September 2016 in Kigoma, as part of a larger evaluation effort of the Project to Reduce Maternal Deaths in Tanzania. The project is a collaboration between Centers for Disease Control and Prevention (CDC) and the Tanzania Ministry of Health, Community Development, Gender, the Elderly and Children (MoHCDGEC), Thamini Uhai, Vital Strategies and EngenderHealth, with financial support from Bloomberg Philanthropies and the H&B Agerup Foundation.
The survey was approved by the CDC Institutional Review Board and the Tanzania National Institute for Medical Research (NIMR) as the main evaluation approach to assess the maternal, child and reproductive health status, health service utilization and behaviors of women ages 15–49 in Kigoma region. Informed consent was obtained from household respondents at the beginning of the household interviews, and separately from eligible respondents at the start of the individual interviews. The consent was given verbally and attested on the paper questionnaire by the interviewer’s signature, date and time of giving consent, which were shown to the respondent, in accordance with the Tanzania NIMR requirements for human subject participation in population surveys. No compensation of any kind was provided to respondents who agreed to voluntarily participate in the survey. A detailed description of the survey methods and procedures are available elsewhere [44].
The survey included a regionally representative probability sample of women ages 15–49 that was selected using the 2012 National Census as the sampling frame. Maps and household listings for each enumeration area were updated during the month prior to the data collection. Trained interviewers obtained informed consent prior to conducting household and individual interviews. If obtained, interviewers then conducted confidential, face-to-face interviews using standardized questionnaires to collect information on households and individual women. The individual questionnaire asked information about a woman’s background characteristics, contraceptive behaviors and use, fertility, and detailed information about the most recent births (i.e., births during January 2014–September 2016) (see Additional file 1).
For the 2016 Kigoma Reproductive Health Survey (RHS), a total of 6461 of 6630 sampled households (97.5%) completed an interview. Within the responding households, 7023 of 7506 women aged 15–49 years (93.6%) responded. Of these women, 3531 (50.1%) women reported at least one live birth between January 2014 and September 2016 and were included in the analysis.
This study derived its main findings from the birth histories, which contained detailed information on birth outcomes (live birth or stillbirth), antenatal care, place and type of delivery and health behaviors during and after pregnancy, including LHU.
Inclusion criteria
For our analyses, we included information on LHU during the last pregnancy and/or labor resulting in live birth for all women who had a live born infant between January 2014 and September 2016. Details on the inclusion criteria and methodology of the survey in general are included in the 2016 Kigoma Region RHS Final Report [44].
Assessment of sociodemographic indicators
The sociodemographic indicators of interests included these selected variables: 1) Age (under 25, 25 to 34, 35 to 49); 2) Marital status (currently in a union, previously in a union, never in a union); 3) Residence (urban, rural); 4) Highest education completed (none, some primary, completed primary and/or higher); and 5) Household wealth index based on household assets (low, middle, high).
Characteristics of pregnancy and delivery
The characteristics of pregnancy and delivery among women reporting LHU were captured through the following variables: 1) LHU during their pregnancy and/or labor (yes/no); 2) Reasons for taking herbs (Induce or sustain labor, treat malaria, treat cold/flu, treat headache, treat convulsions, treat vaginal bleeding, treat stomach pain, for the health of the child, to avoid miscarriage, other (specify)); 3) When LHU was initiated and stopped (1st trimester, 2nd trimester, 3rd trimester, just before delivery, during/after delivery, does not remember); 4) Birth order (continuous); 5) Recommended antenatal care (ANC) received (yes/no). ANC responses were recoded as a dichotomous variable where women either met or did not meet the Tanzanian national guidance recommendation of four or more ANC visits during pregnancy [45]; 5) Gestational age at delivery (continuous); 6) Place of delivery (hospital/health center/dispensary, home, unknown); and 7) Postnatal complications (yes/no): Included only pregnancy obstetric complications with onset during the first 6 weeks of the postnatal period. Postnatal complications included severe bleeding; vaginal discharge, surgical infection, fainting, coma, high fever, pelvic pain, urinary incontinence, and bowel incontinence. Two obstetrician/gynecologist (Ob/Gyn) epidemiologists reviewed the survey responses to include only obstetric complications in the final analysis.
Statistical analysis
Descriptive statistics
We calculated prevalence estimates with 95% confidence intervals for the following selected characteristics: overall LHU, reasons for LHU, LHU by demographic factors and clinically-relevant postnatal obstetric complications by LHU. We also examined the average number of days local herbs were used during pregnancy and/or labor and the proportion of herb users receiving recommended ANC.
Chi-squared statistics
We used chi-squared statistics to assess whether LHU during pregnancy and/or labor varied by residence, age, education level, marital status, wealth, parity, recommended ANC visits, and place of delivery. Also, chi-square tests were performed on whether postnatal obstetric complications varied with LHU during pregnancy and/or labor and by residence, age, education level, marital status, wealth, parity, recommended ANC visits, and place of delivery.
Multivariable models
We constructed two multivariable logistic regression models. The first model examined factors associated with LHU. The second model examined the association between LHU during pregnancy/labor and postnatal obstetric complications, while adjusting for any potential confounders. For both models, only the significant associations (p < 0.05) in the bivariate analyses were included in the full model, which were then removed sequentially based on a threshold of a p < 0.05. For the final multivariable model, we included age in our final model a priori because it is a well-documented risk factor for postnatal complications [46]. The results are presented as adjusted odds ratios (aOR) and 95% confidence intervals (CI). We performed all analyses using SAS® software, Version 9.4 for Windows, using complex survey procedures to account for survey clustering and unequal sampling weights [47].