Health service utilisation lags behind maternal diseases or illnesses during pregnancy in rural south Ethiopia: A prospective cohort study

Maternal survival has improved substantially in the last decades, but evidence on maternal morbidity and health service utilisation for various maternal diseases are scarce in low resource settings. We aimed to measure health service utilisation for maternal illnesses during pregnancy. A cohort study of 794 pregnant women in rural southern Ethiopia was carried-out from May 2017 to July 2018. Disease or illness identification criteria were: symptoms, signs, physical examination, and screening of anaemia. Follow-up was done every two weeks. Data on health service utilisation was obtained from women and confirmed by visiting the health facility. Multilevel, multiple responses, repeated measures, and generalized linear mixed model analysis were used. The cumulative incidence of women experiencing illness episodes was 91%, and there were 1.7 episodes of diseases or illnesses per woman. About 22% of pregnant women were anaemic and 8% hypertensive. Fourteen pregnant women experienced abortions, 6 had vaginal bleeding, 48% pain in the pelvic area, 4% oedema, and 72% tiredness. However, health service utilisation was only 7%. About 94% of anaemic women did not get iron-folic-acid tablet supplementation. Only two mothers with blurred vision and severe headache were referred for further treatment. The main reasons for not using the health services were: the perception that symptoms would heal by themselves (47%), illness to be minor (42%), financial constraints (10%), and lack of trust in health institutions (1%). Risk factors were being older women, poor, having a history of abortion, living far away from the health institution, travelled longer time to reach a health institution, and monthly household expenditure >=30 USD. In Conclusion, there was a high incidence of diseases or illnesses; however health service utilisation was low. Poor understanding of severe and non-severe symptoms was an important reason for low health service utilisation. Therefore, community-based maternal diseases or illness survey could help for early detection. Ministry of Health should promote health education that encourages women to seek appropriate and timely care.

155 women's occupation (housewife=0, others=1), household wealth, and total monthly household 156 expenditure. Individual-level demographic factors were: women's age at first marriage, women's 157 age at first birth, gravidity, parity, birth interval, history of abortion, and history of stillbirth. The 158 community-level factors included their place of residence (Mekonisa=1, Hase-Haro=2, Tumata-159 Chiricha=3), type of road to the nearest health facility (asphalt=0, others=1), walking distance to 160 the nearest health post or to the health centre or to hospital. 161 Outcome variables 162 We used two sources of data to assess health service utilisation and the reasons for not seeking 163 care. For each reported disease or illness episode, the women were asked whether they sought 164 care or not, and then confirmed by visiting the health institutions. We recorded the symptoms of 165 obstetric diseases or illnesses: hypertensive disorders, obstetric haemorrhage, pregnancy-related 166 infection, and other diseases or illnesses. We also recorded symptoms of medical diseases or 167 illnesses: gastrointestinal, psychiatric, and other symptoms. 168 The assessment for pregnancy-related disease or illness measurements were blood pressure and 169 pulse rate (by Riester ri-champion ® N digital apparatus, www.riester.de ), and haemoglobin 170 (HgB) (by HemoCue analyzer ®Hb 301 System, (www.hemocue.com ). The validity and 171 reliability of the measurements were checked regularly before used in the study.
181 Data collection 182 The data were collected prospectively using a pretested, structured, and an interviewer-183 administered questionnaire. The questionnaires were adopted from tools and indicators for 184 maternal and newborn health [31], and from WHO maternal morbidity measurement working 185 group [2]. The questionnaire was initially prepared in English and translated into local language, 186 Gedeo language and Amharic, and back-translated into English. A pre-test was conducted among 187 82 mothers in a neighbouring kebele. The data collectors were trained women, residents of the 188 selected kebeles who could speak the local language (Gedeo language) and Amharic and had 189 completed grade 10. The field nurses and the supervisors were experienced in data collection and 190 supervision. To ensure the data quality, double data entry and validation of data were employed 191 using EpiData version 3.1 and analysed using SPSS version 25 software (SPSS Inc. Chicago,

IL).
193 Quantitative variables 194 The data were assessed using frequency distribution to explore if the variables were normally 195 distributed. We used cross tabulation for categorical variables to see the distribution of each 217 and their community (kebele). Four models were fitted using the GLMM in SPSS version 25.   292 Among diseases that we regarded as severe and needed to be examined by health workers were 293 six women with vaginal bleeding, of whom only one used the health service. Although 22% of 294 pregnant women were anaemic, the uptake of Iron-folic-acid tablet supplementation was low 295 (6%), and the reasons given were that 22 (15.3%) did not have money to visit a health institution, 296 and 6 (4.2%) had no trust in the health institution. Their incidence of diastolic blood pressure 297 was 26.3%; however, the rate of use of health service was only 3.2%. Among 31 (4.3%) women 298 who had oedema, only 13% of them used the health service.
299 Among illnesses that we regarded as non-severe, there was low health service utilisation for 300 tiredness, backache, and dizziness as many women with symptoms perceived their illness to be 301 minor, and they thought they were not important, and some would heal by themselves (Table 3).
302 Only two of the mothers who had a severe headache and excessive tiredness or visual 303 disturbance with a severe headache were referred for further treatment.  . This discrepancy could be explained by the fact 364 that our study was a cohort study with multiple visits to pregnant women homes, and the other 365 studies were cross-sectional. However, the incidence of diseases or illnesses was similar to 366 90.3% from population-based studies in Sri Lanka [37]. The Sri Lankan study showed that 367 maternal illnesses considered as minor were judged not to be minor for women. Though, as 368 shown by our study, the overall rate of use of health services during diseases or illnesses was 369 low. The reasons for not seeking care during diseases or illnesses included a lack of money, or it 370 could be due to longer time to travel to a health facility, or it could be that women who live in