The census collected information from 16,236 women in the reproductive age group. The response rate for the census was 91 %; 7 % of women were not available at the time of data collection and 2 % refused to participate in the study. The LMUP, which was administered to women who were pregnant at the time of census or had a child under two years of age, was completed by 4991 women.
Table 1 describes the socio-demographic and household characteristics of the respondents. The household information included home ownership status, housing construction type, electricity supply, drinking water source, toilet facility, and socioeconomic status. Most women (74 %) lived in permanent housing. Most households (78 %) accessed drinking water at a public or community tap stand and 83 % used a public or shared toilet facility. Mean age of women was 27 (SD 5.05, median 25, IQR 23–30) with number of children ranging from zero to 15 (mean 2.4, SD 1.7, median 2 IQR 1–3). At the time of survey, 23 % women were pregnant, 43 % were in their first postnatal year, and 34 % were in their second. 27 % of women had no formal schooling. Most women (96 %) were not working, 84 % identified as Muslim, and 99 % were either married or cohabitating with their partner.
Acceptability and targeting
There were no missing responses. The full range of LMUP scores was captured (zero to twelve) in the census (Fig. 1). Scores were not normally distributed and had a median of 10 (IQR 8–10) and a mean of 8.6 (SD 2.6). 11 % of women had a score of 0–3 (unplanned), 19 % scored 4–9 (ambivalent), and 70 % scored 10–12 (planned).
Table 2 illustrates women’s responses to individual LMUP questions. Items one (contraception) and six (preconception preparations) showed the least item discrimination. Most participants (94 %) were not using a method of contraception in the month they became pregnant (item one). When looking at preconception preparation (item six), almost all women (96 %) did not take any action to prepare for their pregnancy.
Cronbach’s α for the entire scale was 0.84. Inter-item correlations were all positive. Item-rest correlations were above 0.2 for all items except item six (0.07) (Table 3).
Both construct validity hypotheses were confirmed. Women 30 years of age or above (p = 0.0001), and women with four or more children (p = 0.0001), were more likely to report their pregnancies as unintended (Figs. 2 and 3). The mean LMUP score for women under 30 years was higher (mean 8.8, SD 2.5, median 10, IQR 8–10) than for women above (mean 7.7, SD 3.2, median 10, IQR 6–10). Women with less than four children had a higher mean LMUP score (mean 8.9, SD 2.3, median 10, IQR 9–10) than women with more than three children (mean 7.4, SD 3.3, median 10, IQR 5–10).
Confirmatory factor analysis showed that a one-factor model was a good fit for the data (CFI 0.99, SRMR 0.02). Principal components analysis confirmed that the six items loaded onto one component with eigenvalue 3.5 (Table 3). The KMO was 0.85 and Bartlett’s test for sphericity p < 0.001.
Measurement properties of the LMUP according to time from conception
We did not observe any differences in education level, employment status, or religion across the three groups. The only difference in household characteristics was in home ownership. More than half (59 %) of the pregnant women were in the age group 16–24 years. The proportion was lower for women in their first (50 %) and second (46 %) postnatal years. As expected, women in their first and second postnatal years had more children than pregnant women.
Acceptability and targeting
There were no missing responses. The full range of LMUP scores from zero to twelve was present in the groups of pregnant women and women in their first postnatal year. Women in their second postnatal year had a range of scores from one to twelve.
Cronbach’s α was 0.86 for pregnant women, 0.83 for women in their first postnatal year, and 0.82 for women in their second. Inter-item correlations were positive in all three groups. Item-rest correlations were above 0.2 for items 1–5 in all three groups, except for item one which was borderline (0.19) in the second postnatal year. The item-rest correlation for item six was less than 0.2 in all three groups (Table 4).
Hypothesis testing for each of the three groups confirmed that older women (> 30 years) and women with more children (> 3 children) were more likely to report their pregnancies as unintended. Results were significant for both older women and women with more children (> 3 children) (p = 0.0001 for each of the three groups). The mean LMUP score for pregnant women with less than four children (mean 8.6, SD 2.7, median 8, IQR 8–10) was higher than for women with four or more children (mean 6.2, SD 3.7, median 8, IQR 2–10). Similar results were observed in the first postnatal (mean 9, SD 2.1, median 10, IQR 10–10 for < = 3 children vs. mean 7.4, SD 3.3, median 10, IQR 5–10 for > 3 children) and second postnatal group (mean 9, SD 2.1, median 10, IQR 10–10 for < = 3 children vs. mean 7.4, SD 3.3, median 10, IQR 5–10 for > 3 children). Pregnant women aged < = 30 years had a higher mean LMUP score (mean 8.4, SD 2.8, median 10, IQR 8–10) than older pregnant women (mean 7.3, SD 3.4, median 10, IQR 3–10). Observations were similar for the first postnatal (mean 8.8, SD 2.4, median 10, IQR 8–10 for women < = 30 years vs. mean 7.7, SD 3.2, median 10, IQR 6–10 for women > 30 years) and second postnatal groups (mean 9, SD 2.2, median 10, IQR 10–10 for women < = 30 years vs. mean 7.8, SD 3, median 10, IQR 6–10 for > 30 years).
CFA confirmed the single factor LMUP model as a good fit for the data for every group: pregnant women – CFI 1.0, SRMR 0.02; first postnatal year – CFI 1.0, SRMR 0.01; second postnatal year – CFI 0.99, SRMR 0.04). Principal components analysis showed that all six items loaded onto one component in each group with eigenvalue of 3.6 for pregnant women, 3.4 for women in their first postnatal year, and 3.4 for women in their second. Item six had the lowest component loading in all three groups.
LMUP scores across groups
The median score was 10 (IQR 8–10) for each of the groups. Minor variation was observed in the mean scores of pregnant (8.3, SD 2.9), first (8.6, SD 2.6) and second (8.8, SD 2.4) year postnatal groups (Kruskall Wallis p = 0.002, Jonckheere-Terpstra test p = 0.0002) (Fig. 4).
Multivariable logistic regression showed that the differences in LMUP scores remained after adjusting for parity and age. After controlling for age and parity, women in their first postnatal year (AOR 1.36 (95 %CI 1.16–1.59)) and second postnatal year (AOR 1.63 (95 %CI 1.37–1.92)) were more likely to report their pregnancy as planned than were pregnant women, although the difference between the first and second postnatal year was not significant.