Table 3.

The effect of the MNH+ intervention on the quality of care in government-managed primary healthcare from 2012 to 2016, difference-in-differences analysis

	Control baseline mean or percent	Control follow-up mean or percent	Control diff1	Interv. baseline mean or percent	Interv. follow-up mean or percent	Interv. diff2	β or RR (95% CI)
Processes
Provision of evidence-based care
Routine care (3 items)3	1.75	1.93	0.18	1.90	2.24	0.34	0.16 (−0.03, 0.35)
Basic emergency obstetric and newborn care (6 items)4	2.08	2.42	0.34	2.08	2.58	0.50	0.17 (−1.16, 1.50)
Receipt of services by women
Receipt of IV antibiotic	23.1%	22.9%	−0.2%	18.8%	16.1%	−2.7%	0.86 (0.45, 1.65)
Receipt of uterotonic	75.9%	89.7%	13.8%	82.1%	92.9%	10.8%	0.98 (0.84, 1.12)
Receipt of newborn counseling (6 items)6	4.49	4.46	−0.03	4.25	5.15	0.90	0.74* (0.13, 1.35)
Patient experience and care competence
vNontechnical quality (5 items)7	1.16	1.40	0.24	1.12	1.49	0.37	0.11 (−0.08, 0.30)
vTechnical quality (2 items)8	0.10	0.18	0.08	0.13	0.22	0.09	−0.03 (−0.16, 0.10)
Outcomes
Health outcomes9
Patient is not anemic	-	40.8%	-	-	36.3%	-	0.90 (0.78, 1.05)
Patient is not hypertensive	-	91.7%	-	-	90.9%	-	0.99 (0.97, 1.02)
Maternal mid-upper arm circumference	27.03	28.15	1.12	27.37	28.02	0.65	−0.44 (−0.98, 0.10)
EQ-5D	0.93	0.95	0.02	0.93	0.95	0.02	0.01 (−0.01, 0.03)
Overall quality and satisfaction10
Patient satisfaction with delivery care	47.9%	64.9%	17.0%	47.6%	60.9%	13.3%	0.95 (0.69, 1.30)
Patient perceived quality of delivery care	14.5%	19.1%	4.6%	13.0%	21.2%	8.2%	1.22 (0.58, 2.59)
Provider perceived quality of ANC	15.2%	42.6%	27.4%	27.0%	35.4%	8.4%	0.46 (0.11, 1.87)
Provider perceived quality of labor care	24.3%	35.2%	10.9%	29.7%	44.6%	14.9%	1.04 (0.40, 2.69)
Provider perceived quality of care for obstetric complications	21.2%	18.5%	−2.7%	18.9%	36.9%	18.0%	2.24 (0.66, 7.54)

Notes: Except where noted, all models used generalized estimating equations with an exchangeable correlation structure. For binary quality measures, we used a log link to estimate risk ratios; for continuous measures we used the identity link. The robust sandwich estimator was used to account for clustering at the facility level and a fixed effect for district was included to account for the stratified design. The β coefficients and RR are the difference-in-differences estimates. For example, the increase in number of newborn counseling items received from baseline to endline was 0.74 items higher for women delivering in intervention facilities than women delivering in control facilities.

ANC = Antenatal care

CI = Confidence interval

^* P-value less than 0.05

¹Difference in mean or percentage points between endline and baseline in control group (Control_endline—Control_baseline)

²Difference in mean or percentage points between endline and baseline in intervention group (Intervention_endline—Intervention_baseline)

³Composite indicator using data from facility registers. The summed proportion of deliveries where the infant was breastfed within 1 hour, the baby’s weight was recorded and a partograph was used during delivery.

⁴Composite indicator of six BEmONC signal functions reported by a senior provider to have been performed in the last 3 months: antibiotics administered parenterally, oxytocics administered perenterally, anticonvulsants administered, manual removal of the placenta, removal of retained products, newborn resuscitation.

⁵Women’s report of receipt of three services: provider checked on mother, provider checked on newborn and mother received uterotonic.

⁶Women’s report of receipt of counseling on six items: breastfeeding within the first hour of delivery, breastfeeding exclusively, care of the umbilical cord, need to avoid chilling of baby, immunization and hand washing with soap/water before touching the baby.

⁷Composite indicator of patient reported nontechnical quality. Created from ratings of provider’s explanation, respectful greeting, privacy, facility cleanliness and no disrespectful treatment (values range from 0–5). Count of those with the top rating (e.g. excellent) on Likert scale ranging from poor to excellent. No disrespectful treatment was asked as a yes/no question.

⁸Composite indicator of patient reported technical quality created from ratings of provider knowledge and availability of equipment and medications (values range from 0–2). Count of those with the top rating (e.g. excellent) on Likert scale ranging from poor to excellent.

⁹Comparison of intervention to control at endline and adjusted for age, household wealth (quintiles derived from an 18-question asset index) and district. This association is not causal and can be interpreted as the risk of not having severe anemia is the same in both intervention and control facilities at endline, after adjusting for age, household wealth and district.

¹⁰Quality and satisfaction questions were asked on a Likert scale from poor to excellent or very dissatisfied to very satisfied. Indicators were created to compare those with the top rating (e.g. excellent or very satisfied) to all others.