Advancing the health of women and newborns: predictors of patient satisfaction among women attending antenatal and maternity care in rural Rwanda

Abstract

Objective

Identify predictors of patient satisfaction with antenatal care (ANC) and maternity services in rural Rwanda.

Design

Cross-sectional.

Setting

Twenty-six health facilities in Southern Kayonza (SK) and Kirehe districts.

Participants

Sample of women ≥ 16 years old receiving antenatal and delivery care between November and December 2013.

Intervention

Survey of patient satisfaction with antenatal and delivery care to inform quality improvement (QI) initiatives aimed at reducing neonatal mortality.

Main Outcome Measure

Overall satisfaction with antenatal and delivery care (reported as excellent or very good).

Results

In multivariate logistic regression analysis, high perceived quality [odds ratio (OR) = 3.03, 95% confidence intervals (CI): 1.565.88], respect [OR = 4.13, 95% CI: 2.16–7.89], and confidentiality [SK: OR = 7.50, 95% CI: 2.16–26.01], [Kirehe: OR = 1.54, 95% CI: 0.60–3.94] were associated with higher overall satisfaction with ANC, while having ≥1 child compared to none [OR = 0.46, 95% CI: 0.25–0.84] was associated with lower satisfaction. For maternity services, <5 years of school versus ≥5 years [OR = 0.13, 95% CI: 0.026–0.69] and higher cleanliness [OR = 19.23, 95% CI: 2.22–166.83], self-reported quality [OR = 10.52, 95% CI: 1.81–61.22], communication [OR = 8.78, 95%CI: 1.95–39.59], and confidentiality [OR = 8.66, 95% CI: 1.20–62.64] were all positively associated with high satisfaction. Higher comfort [OR: 0.050, 95% CI: 0.0034–0.71] and Kirehe vs. SK district [OR: 0.21, 95% CI: 0.042–1.01] were associated with lower satisfaction.

Conclusions

Patient-centeredness (including interpersonal relationships), organizational factors, and location are important individual determinants of satisfaction for women seeking maternal care at study facilities. Understanding variation in these factors should inform QI efforts in maternal and newborn health programs.

Introduction

While there has been recent progress towards reducing maternal and neonatal mortality (MDG 5), critical gaps remain in the developing world, with more than 300 000 women dying each year during pregnancy and childbirth [1], and almost 3 million neonates dying each year [2]. Quality antenatal care (ANC) and facility-based delivery with a skilled health worker are important interventions to improve pregnancy outcomes, reducing maternal and perinatal mortality [3, 4], yet coverage of these services remains insufficient. In Rwanda, reduction of maternal and neonatal mortality is a key national priority [5]. As of 2015, although 99% of pregnant women received at least one ANC visit from a skilled provider, less than half of Rwandan women report receiving all four recommended standard ANC visits [6].

Patient satisfaction can influence health-seeking behaviors and adherence to care, including in ANC and maternity settings in Africa [7–10]. Satisfied patients are more likely to return to care and recommend services to others, which may support effective coverage of services [10, 11]. Prior research suggests that in Sub-Saharan Africa, factors that influence patients’ satisfaction with health services include interpersonal (provider–patient) relationships, technical quality and organizational factors (such as cleanliness and comfort) [10–14].

The Rwandan Ministry of Health (MOH), supported by Partners In Health (PIH), launched the All Babies Count Initiative (ABC) in 2013 [15]. The goal of ABC was to reduce neonatal mortality by improving the quality of ANC and maternity care in two rural districts over an 18-month period. ABC consisted of three components: (a) healthcare worker training followed by on-site mentorship, (b) essential equipment and materials and (c) a quality improvement (QI) strategy using district-wide learning collaboratives, which included training and support for improving patient-centered care.

In order to assess experiential quality as a key domain of quality and a likely driver of care-seeking behavior, a patient satisfaction survey was conducted at baseline. The objective of this study is to describe baseline levels and primary predictors of patient satisfaction with antenatal and maternity care services at the time of delivery in rural Rwanda and to inform the QI processes of the ABC initiative.

Methods

Study setting

The study was conducted in all 16 health centers (HCs) in Kirehe and 8 HCs in Southern Kayonza (SK) districts and the two district hospitals. ANC occurs at the HCs and deliveries at all 26 facilities. These rural districts are located in the Eastern Province of Rwanda, serving a catchment population of ~480 000 [16]. In partnership with the MOH, PIH has provided clinical and infrastructure support to the public health system in SK since 2005, expanding to Kirehe in 2009 [17, 18]. The ABC initiative built on an ongoing district-wide strengthening program started in 2009 to reduce neonatal mortality [17].

Survey design

A conceptual framework was developed to inform survey design. Using a model created by Donabedian [19], three domains of factors influencing patient satisfaction were identified: (1) organizational components, (2) technical quality and (3) interpersonal factors. These factors have been found to influence patient satisfaction with maternal health services in Sub-Saharan Africa [11, 20, 21]. Our conceptual framework outlines the link between these factors and satisfaction with antenatal and maternity care (Fig. 1). Using this framework, 22 independent variables with potential links to patient satisfaction were identified and included in the survey: (1) patient characteristics, including age, marital status, education level, socioeconomic status, number of children, patient knowledge of danger signs and access (measured by travel time, cost of transportation, difficulty of payment for transport and overall difficulty of obtaining transportation); (2) organizational components, including length of wait time, acceptability of wait time, cleanliness and comfort; (3) reported technical quality of care received, including technical skills and perceived quality of care and (4) interpersonal factors, including understanding, communication, respect, promptness and confidentiality. The primary outcome—overall satisfaction—was assessed through the question: ‘Overall, how satisfied were you with the care that you received?’ Additional survey questions can be found in the Appendix.

Figure 1.

Conceptual framework for antenatal care (ANC) and maternity service satisfaction.

Questions regarding satisfaction and perceptions of care employed a standard 5-point likert scale: excellent, very good, good, fair and poor. Both antenatal and maternity surveys were translated into Kinyarwanda, back-translated to ensure accuracy and conducted in Kinyarwanda.

Surveys were administered by four data collectors, who received training in survey administration, data collection and research ethics. Each participant received a written and verbal explanation of purposes, risks, benefits and alternatives to study participation, and provided informed consent (signed or marked if unable to write). No compensation was provided for participation in the study.

Data collectors visited health facilities on ANC visit days between November and December 2013, with all women attending ANC eligible for study participation. Additionally, any woman in the maternity ward who had delivered at least 4 h previously was eligible to participate in the maternity survey. Additional inclusion criteria included being at least 16 years of age and willingness and ability to provide consent. Exclusion criteria included acute maternal illness such as clinical instability, and having had a fetal or neonatal death in the current delivery. All surveys were conducted in a private location away from health providers and other patients. The study was approved by the Rwanda National Ethics Committee and the Partners Institutional Review Board.

Statistical methods

Potential predictors of satisfaction were converted to binary variables. For questions measured on a likert scale (technical skills, quality of care, understanding, communication, respect, promptness, confidentiality and overall satisfaction), responses were divided into positive responses (excellent or very good: EVG) or neutral/negative (good, fair or poor). Our primary outcome was high (EVG) overall satisfaction.

Earlier studies in this population revealed differences in measures of quality of care across districts [15], so we included district as a potential effect modifier. Bivariate associations were tested between reported EVG satisfaction and the independent predictors using chi-squared test. Breslow–Day tests of homogeneity were used to test for significant effect measure modification by district of the association between independent predictors and overall patient satisfaction. All variables found to be statistically significant (P < 0.05) in the bivariate analysis were kept for use in regression analysis, except for ‘difficulty obtaining transport’ due to substantial missing responses for this question across surveys.

Multivariate logistic regression analysis was used to identify predictors of EVG satisfaction. A backwards, stepwise selection approach was used to select covariates with a threshold of P < 0.05, except for district, which was included in the final model.

After fitting the main effects models, interaction terms were tested between district and independent predictors found to have significantly different stratified odds ratios as identified by the Breslow–Day tests, and any interaction terms that were significant were also included in the final model. For goodness-of-fit, calibration using Hosmer–Lemeshow chi-square tests and discrimination using c-statistics were assessed. Statistical analysis was done using STATA version 13 (StataCorp.2013, Stata Statistical Software: Release 13. College station, TX: StataCorp LP). Results are reported as Odds Ratios (OR), 95% confidence intervals (CI) and P-values with alpha = 0.05.

Results

ANC and maternity demographic characteristics

This study included 278 women from Kirehe district (ANC: 204, maternity: 74) and 198 from SK (ANC: 166, maternity: 32). Demographics were similar across districts for both surveys (Table 1). Overall, about half of women reported high satisfaction with services, (ANC: 59%, maternity: 52%). Patients in SK reported higher satisfaction than patients in Kirehe (ANC: 83% vs. 40%, maternity: 84% vs. 39%).

Table 1.

Demographic characteristics of patients who attended antenatal care and maternity services

Characteristics	ANCa						Maternity
	Kirehe		Kayonza		Total		Kirehe		Kayonza		Total
	n	%	n	%	N	%	n	%	n	%	N	%
Total	204	100	166	100	370	100	74	100	32	100	106	100
Personal information
Age
Mean, SDb	26.6	6.0	27.1	6.6	26.8	6.3	26.6	6.2	26.8	7.5	26.7	6.6
Current marital status
Married	103	50	78	47	181	49	33	45	17	53	50	47
Not married	101	50	87	53	188	51	41	55	15	47	56	53
How many years of school have you completed?
Up to 4 years primary	116	57	59	38	175	49	46	62	15	54	61	60
5 years primary or more	88	43	95	62	183	51	28	38	13	46	41	40
Which of the following items are present at your home?
Radio	131	64	121	73	252	68	43	59	19	61	62	60
Bicycle	91	45	70	42	161	44	33	45	20	65	53	51
Electricity	23	11	35	21	58	16	4	5	4	13	8	8
Goats	90	44	79	48	169	46	37	51	16	52	53	51
Running water	11	5	7	4	18	5	9	12	3	10	12	12
Motobike	4	2	5	3	9	2	5	7	1	3	6	6
Mobile phone	120	59	121	73	241	65	32	44	16	52	48	46
Cows	68	33	45	27	113	31	21	29	7	23	28	27
How many living children do you have, including your newborn?
Median, IQRc	1	[0–3]	1	[0–3]	1	[0–3]	2	[1–4]	2	[1–4]	2	[1–4]
Healthcare services
Did you receive antenatal care at this facility where you delivered your baby?
Yes							58	78	28	88	86	81
No							15	20	4	13	19	18
Did not receive any care before delivery							1	1	0	0	1	1
Number of ANC visits
1	130	64	91	55	221	60	7	10	1	3	8	8
2	47	23	41	25	88	24	19	26	5	16	24	23
3	16	8	22	13	38	10	34	47	14	44	48	46
4	6	3	11	7	17	5	12	17	10	31	22	21
>4	5	2	0	0	5	1	0	0	2	6	2	2
I do not remember but I had more than 1 visit	0	0	1	1	1	0
Overall satisfaction (EVGd)	81	40	134	83	215	59	28	39	26	84	54	52

^aANC, antenatal care.

^bSD, standard deviation.

^cIQR, interquartile range.

^dEVG, excellent very good.

Bivariate analysis

Among ANC respondents, patient, organizational, technical and interpersonal relationship factors were associated with overall satisfaction in our bivariate analyses, although differed between the districts (Table 2a). Lower parity (no prior children versus ≥1 children) was associated with greater satisfaction (Kirehe: 53% vs. 34%, P = 0.010, SK: 89% vs. 80%, P = 0.162). Patients reporting very clean (Kirehe: 79% vs. 37%, P = 0.002, SK: 92% vs. 80%, P = 0.069) or very comfortable (Kirehe: 50% vs. 39%, P = 0.504, SK: 92% vs. 77%, P = 0.019) facilities reported higher satisfaction than those who did not. Patients who reported EVG technical skills (Kirehe: 52% vs. 38%, P = 0.190, SK: 96% vs. 65%, P < 0.001) and quality of care (Kirehe: 66% vs. 32%, P < 0.001, SK: 95% vs. 49%, P < 0.001) reported higher satisfaction compared to those who did not. Several interpersonal relationship factors tested were also significantly associated with high satisfaction in both districts, including higher ratings of provider understanding, (Kirehe: P = 0.001, SK: P < 0.001), communication (Kirehe: P = 0.048, SK: P < 0.001), respect (Kirehe: P < 0.001, SK: P < 0.001), promptness (Kirehe: P = 0.001, SK: P = 0.039), and confidentiality (Table 2a).

Table 2a.

Bivariate relationships between demographic and health system factors associated with high patient satisfaction in antenatal services

Characteristics	Kirehe		Southern Kayonza		Total		BD
	n (%)	p-valuea	n (%)	p-valuea	n (%)	p-valuea	p-valueb
Patient characteristics
Age range		0.185		0.012		0.053	0.2357
16-29	59 (43)		91 (88)		150 (63)
30+	22 (33)		43 (73)		65 (52)
Marital status		0.344		0.055		0.634	0.0344
Married	37 (37)		74 (88)		111 (60)
Unmarried	44 (43)		59 (77)		103 (58)
Education level		0.404		0.446		0.014	0.8684
Up to 4 years primary	43 (37)		45 (79)		88 (51)
5+ years primary/secondary	38 (43)		78 (84)		116 (64)
Socioeconomic status (total items at home)		0.286		0.76		0.219	0.3957
2 or fewer items	33 (36)		52 (84)		85 (55)
3+ items	48 (43)		82 (82)		130 (62)
Number of children		0.010		0.162		0.012	0.8720
0 children	36 (53)		47 (89)		83 (69)
1+ children	45 (34)		87 (80)		132 (55)
Travel time		0.574		0.815		1.000	0.6267
<1 hour	20 (54)		28 (80)		48 (67)
>1 hour	7 (64)		3 (75)		10 (67)
Travel payment		0.656		0.322		0.945	0.2844
Free	14 (58)		19 (76)		33 (67)
Paid	13 (52)		15 (88)		28 (67)
Travel payment difficulty		0.662		0.136		0.241	0.2361
A little/very difficult	7 (58)		10 (100)		17 (77)
Not difficult	8 (50)		8 (80)		16 (62)
Travel difficulty		0.171		0.018		0.009	0.2484
A little/very difficult	45 (57)		74 (93)		119 (75)
Not difficult	17 (44)		22 (76)		39 (57)
Patient knowledge		0.512		0.991		0.003	0.5742
7-8 correct out of 8	3 (30)		48 (83)		51 (75)
6 or fewer correct	78 (40)		86 (83)		164 (55)
Organizational factors
Wait time		0.999		0.918		0.070	0.9310
<1 hour	41 (40)		97 (83)		138 (63)
>1 hour	39 (40)		37 (82)		76 (53)
Wait okay		0.178		0.735		0.839	0.3116
Acceptable	55 (37)		109 (83)		164 (59)
Too long/Much too long	25 (48)		25 (81)		50 (60)
Cleanliness		0.002		0.069		<0.0001	0.4397
Very clean	11 (79)		36 (92)		47 (89)
Other	70 (37)		98 (80)		168 (54)
Comfortable		0.504		0.019		<0.0001	0.3666
Very comfortable	5 (50)		55 (92)		60 (86)
Other	76 (39)		78 (77)		154 (52)
Technical adequacy
Technical skills		0.190		<0.0001		<0.0001	0.0076
Excellent/Very good	12 (52)		85 (96)		97 (87)
Other	68 (38)		40 (65)		108 (45)
Overall quality of care		<0.0001		<0.0001		<0.0001	0.0135
Excellent/Very good	31 (66)		109 (95)		140 (86)
Other	48 (32)		20 (49)		68 (36)
Interpersonal relationship
Understanding		0.001		<0.0001		<0.0001	0.1879
Excellent/Very good	26 (63)		106 (91)		132 (84)
Other	55 (34)		27 (60)		82 (40)
Communication		0.048		<0.0001		<0.0001	0.0036
Excellent/Very good	17 (57)		105 (95)		122 (87)
Other	64 (37)		28 (56)		92 (41)
Respect		<0.0001		<0.0001		<0.0001	0.0062
Excellent/Very good	30 (63)		110 (95)		140 (85)
Other	51 (33)		23 (51)		74 (37)
Promptness		0.001		0.039		<0.0001	0.7413
Excellent/Very good	26 (62)		61 (90)		87 (79)
Other	55 (34)		71 (77)		126 (50)
Confidentiality		0.015		<0.0001		<0.0001	0.0105
Excellent/Very good	17 (61)		120 (92)		137 (86)
Other	64 (37)		14 (45)		78 (38)

Bold text: p-values<0.05.

^aChi-square test p-value.

^bBreslow-Day test of homogeneity for stratified odds ratios.

For maternity services, patients reporting higher cleanliness (Kirehe: 75% vs. 31%, P = 0.004, SK: 100% vs. 72%, P = 0.046), and higher provider technical skills (Kirehe: 71% vs. 31%, P = 0.005, SK: 100% vs. 64%, P = 0.007) had higher overall satisfaction compared to those who did not. The same pattern was seen for women reporting EVG overall quality of care (Kirehe: 56% vs. 33%, P = 0.097, SK: 100% vs. 56%, P = 0.001). All interpersonal relationship factors, were positively associated with high satisfaction, including communication, promptness and confidentiality (Table 2b).

Table 2b.

Bivariate relationships between demographic and health system factors associated with high patient satisfaction with maternity services

Characteristics	Kirehe		Southern Kayonza		Total		BD
	n (%)	p-valuea	n (%)	p-valuea	n (%)	p-valuea	p-valueb
Patient characteristics
Age range		0.382		0.686		0.328	0.9661
16-29	20 (43)		18 (86)		38 (56)
30+	8 (32)		8 (80)		16 (46)
Marital status		0.936		0.682		0.902	0.6868
Married	15 (39)		13 (87)		28 (52)
Unmarried	13 (39)		13 (81)		26 (53)
Education level		0.025		0.278		0.367	0.0432
Up to 4 years primary	22 (49)		14 (78)		36 (57)
5+ years primary/secondary	6 (22)		12 (92)		18 (45)
Socioeconomic status (total items at home)		0.600		0.743		0.608	0.9505
2 or fewer items	14 (41)		12 (86)		26 (54)
3+ items	13 (35)		13 (81)		26 (49)
Number of children		0.212		0.818		0.232	0.7146
First child	8 (30)		9 (82)		17 (45)
Has prior children	20 (44)		17 (85)		37 (57)
Travel time		0.543		0.558		0.338	0.4845
<1 hour	24 (43)		23 (85)		47 (57)
>1 hour	4 (33)		2 (100)		6 (43)
Travel payment		0.278		0.671		0.515	0.2979
Free	0 (0)		1 (100)		1 (33)
Paid	21 (38)		22 (85)		43 (52)
Travel payment difficulty		0.175		0.108		0.744	0.0416
A little/very difficult	15 (43)		11 (73)		26 (52)
Not difficult	4 (24)		8 (100)		12 (48)
Travel difficulty		0.672		0.277		0.937	0.2450
A little/very difficult	12 (46)		10 (77)		22 (56)
Not difficult	13 (41)		12 (92)		25 (56)
Patient knowledge		0.119		0.060		0.918	0.0160
6 correct out of 6	18 (47)		9 (69)		27 (53)
<6 correct	10 (29)		17 (94)		27 (52)
Organizational factors
Transfer time		0.659		none		0.580	1.000
<1 hour	1 (33)		6 (100)		7 (78)
>1 hour	2 (50)		1 (100)		3 (60)
Cleanliness		0.004		0.046		<0.0001	0.407
Very clean	9 (75)		12 (100)		21 (88)
Other	18 (31)		13 (72)		31 (40)
Comfortable		0.064		0.070		0.182	0.0097
Very comfortable	0 (0)		11 (100)		11 (69)
Other	28 (42)		15 (75)		43 (49)
Technical adequacy
Technical skills		0.005		0.007		<0.0001	0.2237
Excellent/Very good	10 (71)		17 (100)		27 (87)
Other	18 (31)		9 (64)		27 (38)
Overall quality of care		0.097		0.001		<0.0001	0.0449
Excellent/Very good	10 (56)		20 (100)		30 (79)
Other	17 (33)		5 (56)		22 (37)
Interpersonal relationship
Communication		0.004		0.004		<0.0001	0.1498
Excellent/Very good	14 (64)		18 (100)		32 (80)
Other	14 (28)		8 (62)		22 (35)
Respect		0.002		0.038		<0.0001	0.6852
Excellent/Very good	14 (67)		18 (95)		32 (80)
Other	14 (28)		8 (67)		22 (35)
Promptness		0.028		0.056		<0.0001	0.6246
Excellent/Very good	10 (63)		21 (91)		31 (79)
Other	18 (32)		5 (63)		23 (36)
Confidentiality		0.002		0.005		<0.0001	0.6634
Excellent/Very good	9 (82)		19 (95)		28 (90)
Other	19 (31)		4 (50)		23 (33)

Bold text: p-values<0.05.

^aChi-square test p-value.

^bBreslow-Day test of homogeneity for odds ratios stratified by district.

Multivariable analysis—logistic regression

While many significant bivariate associations with ANC satisfaction were identified, fewer remained upon adjustment for other covariates in our ANC regression model (Table 3). In our final ANC model, we found that patients with more living children (1+ vs. 0) were less likely to report high satisfaction [OR = 0.46, 95% CI: 0.25–0.84]. Women who reported greater perceived quality of services [OR = 3.03, 95% CI: 1.56–5.88] and higher respect [OR = 4.13, 95% CI: 2.16–7.89] were more likely to report high overall satisfaction. In SK, patients who reported high confidentiality had 7.5 times the odds of reporting high overall satisfaction compared to those reporting low confidentiality [OR = 7.50, 95% CI: 2.16–26.01], while in Kirehe, this association was not significant [OR = 1.54, 95% CI: 0.60–3.94].

Table 3.

Multivariate logistic regression model^a with predictors of reported ‘EVG’ satisfaction^b among antenatal care respondents

Reported EVG overall patient satisfaction	Odds ratio	95% Conf. interval	P>|z|
Number of live children (1+ vs. 0)	0.46	0.25–0.84	0.012
EVG quality vs. other	3.03	1.56–5.88	0.001
Kayonza: EVG confidentiality vs. other	7.50	2.16–26.01	0.002
Kirehe: EVG confidentiality vs. other	1.54	0.60–3.94	0.37
EVG respect vs. other	4.13	2.16–7.89	<0.0001
Kirehe vs. Kayonza	1.08	0.43–2.74	0.87

^aHosmer–Lemeshow chi-squared test P-value = 0.5828, c-statistic = 0.8446.

^bEVG, excellent or very good.

In our maternity regression model, several variables across domains were found to be significant predictors of high overall satisfaction. Our final model (Table 4) showed that participants with higher levels of education (≥5 years of school) were less likely to report high overall satisfaction than those with lower levels of education [OR = 0.13, 95% CI: 0.03–0.69]. Reporting high quality of maternity services [OR = 10.52, 95% CI: 1.81–61.22], positive nurse communication [OR = 8.78, 95% CI: 1.95–39.59], and positive confidentiality practices [OR = 8.66, 95% CI: 1.19–62.64] were all associated with high overall satisfaction in our model. While higher ratings of cleanliness were associated with higher overall satisfaction [OR = 19.23, 95% CI: 2.22–166.83]; higher ratings of comfort were significantly associated with lower likelihood of reporting high satisfaction [OR = 0.049, 95% CI: 0.0034–0.71]. Receiving care in Kirehe district was associated with significantly lower odds of high satisfaction compared to SK district [OR = 0.21, 95% CI: 0.04–1.03].

Table 4.

Multivariate logistic regression model^a with predictors of reported ‘EVG’ satisfaction^b among maternity respondents

Reported EVG overall patient satisfaction	Odds ratio	95% Conf. interval	P > |z|
5+ years school vs. ≤4 years primary school	0.13	0.026–0.69	0.016
EVG cleanliness vs. other	19.23	2.22–166.83	0.007
EVG comfort vs. other	0.049	0.0034–0.71	0.027
EVG quality vs. other	10.52	1.81–61.22	0.009
EVG communication vs. other	8.78	1.95–39.59	0.005
EVG confidentiality vs. other	8.66	1.20–62.64	0.032
Kirehe vs. Kayonza	0.21	0.042–1.01	0.052

^aHosmer–Lemeshow chi-squared test P-value = 0.7163, c-statistic = 0.9182.

^bEVG, excellent or very good.

Discussion

Our study assessed women’s satisfaction with ANC and maternity services in rural Rwanda prior to initiation of an intervention to improve service quality and reduce maternal and neonatal mortality. Our findings showed that about half of ANC and maternity patients were highly satisfied, thus suggesting substantial room for improvement. For ANC patients, no prior children, patient–provider interactions, including respect and confidentiality as well as perceived technical quality were all associated with higher satisfaction. Among women in maternity, while interpersonal interactions and quality were still important, organizational factors including cleanliness and comfort, as well as education level, were also associated with patient satisfaction.

Our results provide insight into the drivers of patient satisfaction with maternal care in Rwanda at the patient and care delivery levels. Our finding that women with no prior living children reported greater satisfaction with ANC suggests that women’s expectations from prior birth experiences may influence satisfaction, a finding that is consistent with prior studies [22]. Additionally, while we found that lower levels of maternal education were associated with greater satisfaction with maternity care, prior research has shown mixed results on the relationship between patient education and satisfaction with care. For example, a study in Ethiopia found greater satisfaction with ANC among women with lower levels of education [23], while a Ugandan study found that women with some education reported higher satisfaction with health services compared to women with no education [14].

Consistent with findings from other studies [10–13], interpersonal relationships were important contributors to patient satisfaction with ANC and maternity care in our models, specifically respect, communication, and confidentiality. Our finding that cleanliness was a significant predictor of satisfaction with maternity care is also consistent with prior research indicating that this is an important organizational factor contributing to patient satisfaction [13], although recent research has suggested that this may be relatively less important to satisfaction than technical and interpersonal components of care [20]. Surprisingly, we found a small inverse association between reported comfort and reported satisfaction in the maternity model. In the bivariate analysis, we found that no Kirehe patients who reported high comfort also reported high satisfaction, and no SK patients who reported high comfort also reported low satisfaction, which produced uninformative cells in the district-level comparisons. For this reason, we would be cautious in generalizing the finding on comfort beyond our sample.

Our data showed a significant difference in satisfaction between the two districts studied. This may reflect true differences in the quality and experience of care provided, as suggested in a prior study reporting technical measures of quality in the region [15]. Between the districts, SK has received PIH support for a longer period of time than Kirehe, which may have contributed to increased experiential quality of services. Furthermore, the greater geographic spread of health centers in Kirehe as compared to SK may have resulted in greater systems-level challenges to the provision of responsive and high-quality health services in Kirehe.

There are some limitations to consider in the interpretation of this study. Our analysis is based on cross-sectional data, limiting the ability for causal inference. Additionally, our study was powered for ANC outcomes, so the number of maternity surveys collected was small, which may limit statistical power and generalizability of predictors of maternity services satisfaction to other contexts.

Another possible limitation is a bias towards reporting positive satisfaction if deemed more socially desirable [24]. However, we found that many participants provided answers that were not at the higher end of the satisfaction spectrum, indicating that any such desirability effect is likely small. Since mothers who suffered poor outcomes (maternal death, still births and neonatal death) were excluded from maternity surveys, maternity satisfaction results could in fact be biased towards more positive reported satisfaction. However, as these poor outcomes are relatively rare, and near-miss events are typically more common and included in these surveys, we expect this bias to be minimal.

Finally, although we assessed patients’ perception of quality of care provided, we did not have an objective measure of technical quality of care provided at the time of assessment, nor did we measure patients’ actual health status, or their expectations for the care they received. Studies have suggested that a patient’s general health level and the actual outcome of care can influence satisfaction [13, 25], suggesting that this could be an important measure for future research.

Patient satisfaction is increasingly recognized as an essential component of evaluating health systems in developing countries [26]. While a few studies have found significant effects of health interventions in Africa on improving patient satisfaction [27–30], patient satisfaction is still infrequently measured in assessing the success of health interventions in resource-limited settings. This study demonstrates both the feasibility and utility of measuring patient satisfaction as a critical component of QI initiatives. As the world moves towards more people-centered medical care, efforts are underway to make medical service delivery more responsive to individual preferences, needs and values. At the completion of this survey, we presented the results in district-wide coordination meetings, allowing participants to discuss shortcomings in satisfaction measures and ideas for QI.

In promoting the health of pregnant women and newborns, ensuring both access to and utilization of high-quality people-centered ANC and maternity care is essential. Research suggests that patients who are more satisfied with the care they receive may be more likely to seek ANC and maternity care [8, 9]. Our results underscore that satisfaction with maternal care can vary by location and individual, reinforcing the need for healthcare delivery interventions to reflect local needs and context and highlighting the importance of patient-centered metrics in assessing quality of care. Only through this integrated measurement and local response to identified gaps can we achieve our goals of improving quality and increasing adherence to ANC and maternity care.

Funding

This work was supported by the Doris Duke Charitable Foundation’s Africa Health Initiative. S.H.I. was supported in part by the National Institutes of Health research training grant [NIH, T32 CA 009001].