The effect of off-hour delivery on severe maternal morbidity: a population-based cohort study

Abstract

Background

Severe maternal morbidity is an indicator for quality of maternal care. Recently, there has been growing interest in identifying which provision factors affect the quality of maternity care. The extent to off-hour delivery on SMM rates contributes to individual or provision factor in Korea has not been studied. This study aimed to determine the relationship between off-hour delivery and SMM during childbirth hospitalization.

Methods

This is a population-based retrospective cohort study. Data were extracted from the Korean National Health Insurance Service-National Sample Cohort for 90 072 cases of delivery in Korea between 2003 and 2013. The main outcome was SMM which was determined using the Center for Disease Control and Prevention’s algorithm. A generalized estimating equation model with log link was performed for the relationship with SMM and day/time of delivery adjusted covariates.

Results

Of the 90 072 delivery cases, 2085 (2.31%) had SMM. Women who were on weekdays at night time or on weekend delivery had a higher risk of SMM compared with those who were on weekdays at daytime (RR 1.26, 95% CI 1.10–1.46, and RR 1.58, 95% CI 1.30–1.93, respectively).

Conclusion

Weekday at night time or weekend delivery was related to the risk of SMM. Policymakers should provide financial support and systematically allocate adequate human resources and labour facilities in vulnerable areas, as well as during weekends and night times to improve the quality of intrapartum and postpartum maternity care.

Introduction

Maternal mortality has been the typical indicator of the quality of maternal care. The United Nations achieved a 45% reduction in the maternal mortality ratio (MMR) from 1990 to 2015 through global efforts to accomplish the Millennium Development Goal,^1,2 and have re-affirmed the reduction of maternal mortality as a global priority in the coming decades through adoption of the Sustainable Development Goals.³ In spite of these global activities, the burden of high maternal death is still a relevant public health problem of under developed, developing and developed countries.⁴ In Korea, though most expectant mothers had prenatal care and gave birth in the hospital with obstetricians, MMR was reported to be over 11 per 100 000 live births until in 2014, and was the third of the highest MMR among Organization for Economic Cooperation and Development countries in 2015.⁵ However, there is less evidence of the association between quality factors and maternal death in Korea.⁶ To end preventable maternal adverse outcome during childbirth or postpartum period, alternative indicators for maternal death are needed.

Severe maternal morbidity (SMM) is a more useful indicator for evaluation and improvement of maternal health services than is the MMR⁷ and has grown in use as an indicator of the quality of obstetrical care. SMM can be defined as unintended outcomes of the process of labour and delivery that result in significant short-term or long-term consequences to a woman’s health.⁸ Previous research has identified a number of individual risk factors for SMM including employment status,⁹ household income¹⁰ and residence in disadvantaged areas.^11,12 Additionally, obstetric history or performance factors have been associated with SMM including maternal age,^11,13,14 comorbidities,¹⁵ obstetric complications,^13,16 multiple births¹⁷ and caesarean section delivery.^11,13,15 Furthermore, research has also reported provision factors to be related to SMM, including incomplete or inappropriate treatment¹⁸ and hospital volume.¹⁶ However, there has been less research focused on the relationship between SMM and quality factors of healthcare system such as staffing levels or available resources.

Recently, there has been growing interest in identifying which provision factors affect the quality of patient care and which quality outcome might improve in processes of care.¹⁹ In particular, working hours, off-hours shifts and shift lengths have been related to medical errors and poor outcomes because of staffing’s fatigue or sleep deprivation.^20,21 Thus, several European and US studies have documented differences in the complication rates between weekdays and weekends in obstetric units^19,22 and gaps in maternal morbidity between the time of birth.²³ However, there are less shown that the relationship between off-hour childbirth (night time, weekends or holidays) and SMM adjusted individuals, obstetrics and provisions factors using nationally representative data. Therefore, this analysis aimed to explore the relationship between day and time of delivery and SMM during childbirth hospitalization in Korea.

Methods

Data source

The Korean National Health Insurance Service-National Sample Cohort (NHIS-NSC) is the national representative cohort database for the 2002–13 period in South Korea, which includes information on ∼1 million Koreans since 2002. The NHIS-NSC aims to track patient and clinical characteristics over time, reveal epidemiological causes of diseases and develop health policies. The NHIS-NSC used a 2.5% (n = 1 025 340) stratified random sampling method, including age, sex, residence, health insurance type, family income decile and individual total medical costs in 2002. Data do not contain direct personal identifiers but include the unique de-identified numbers of the patients, age, sex, type of insurance, diagnoses according to the International Classification of Diseases, 10th Revision (ICD-10), medical costs, procedures and prescribed drugs. In addition, the unique de-identified numbers were linked to mortality information from the Korean National Statistical Office.²⁴

Diagnosis and procedure codes were identified for all women aged 15 years or older and <50 years old, who had a delivery hospitalization, and were continuously enrolled for at least 1 year before delivery through 42 days after delivery between 1 January 2003 and 19 November 2013. Deliveries were defined as any inpatient hospital admission records including a pregnancy-related diagnosis or procedure code for vaginal or caesarean delivery among women whose pregnancies had reached full term (i.e. ≥37 weeks gestational age). Total selected participants were 90 072 delivery cases between 2003 and 2013.

Severe maternal morbidity

The outcome variable was SMM. SMM was identified by having at least 1 of the 25 previously established ICD-9-CM diagnosis and procedure codes (SMM indicators) during delivery hospitalizations using an algorithm developed by researchers at the Centers for Disease Control and Prevention (CDC).^25,26 The algorithm identified 25 indicators of SMM that represented either serious complications of pregnancy or delivery. Of the 25 indicators, 18 were identified using diagnosis codes converted from ICD-9-CM to ICD-10 and seven indicators used procedure codes. In addition, the delivery hospitalization had to meet at least one of the following two criteria to be considered SMM: (i) the mother was admitted in the ICU; or (ii) the mother died during delivery hospitalization.^17,27

Covariates

Day and time of delivery were identified from the Electronic Data Interchange (EDI) codes. The data contained delivery day which was classified as being: weekday, weekday night time and weekend and holiday. Therefore, this study identified three categories as follows: weekday and daytime delivery, weekday and night-time delivery and weekend/holiday delivery.

Maternity factors included maternal age (15–49 years old), household income level (quintile), residential area (metropolitan/city/rural), type of insurance (self-employed insured/employee insured/medical aid) and working status (working/not working). Adequate prenatal care was estimated by the Kessner Adequacy of Prenatal Care Index:²⁸ mode of delivery (spontaneous vaginal delivery/instrumental delivery/caesarean section delivery), parity (1/2/3+) and twin birth status (singleton/twin birth) were included in this analysis. In this study, comorbidities were estimated by the Howell’s study.²⁹ Provision factors included the type of hospital with regard to the number of beds (clinic/hospital/general hospital), hospital ownership (government/private or for-profit), teaching status (not teaching/teaching) and delivery year.

Statistical analysis

We calculated the distribution of the general characteristics of the study participants who delivered between 2003 and 2013. The characteristics of the SMM and non-SMM groups were compared using Pearson chi-square tests. We confirmed the strength of association in the SMM within independent variables using Phi and Cramer’s V coefficients. Using a generalized estimating equation (GEE) model with log link and a first-order autoregressive correlation structure, we estimated adjusted risk ratios (RR) and 95% confidence intervals (CIs) to examine which maternal characteristics were associated with SMM adjusted confounders. A GEE model with log link was also performed for sub-groups analysis by SMM and by day/time of delivery. All statistical analyses were performed using SAS 9.4 (SAS Institute, Inc., Cary, NC, USA). The level of significance was set at P < 0.05.

This study adhered to the tenets of the Declaration of Helsinki, and the study design was reviewed and approved by the Yonsei University Health System, Institutional Review Board (Y-2017-0002).

Results

Of the 90 072 women included in this sample, 2085 (2.31%) experienced SMM during the delivery hospitalization (table 1). Regarding time and day of delivery, the proportion of deliveries on weekends or holidays was the highest for those with SMM, followed by weekdays night time and weekdays daytime (3.57%, 2.82% and 2.37%, respectively).

Table 1

General characteristics of the study sample who gave birth between 2003 and 2013

	Severe maternal morbidity
	Total (N = 90 072)		No (N = 87 987)		Yes (N = 2085)
	N	(%)	N	(%)	N	(%)	P value
Day/time of delivery
Weekday daytime	78 695	(87.37)	76 924	(97.75)	1771	(2.25)	0.0003
Weekday night time	8381	(9.30)	8165	(97.42)	216	(2.58)
Weekend or holiday	2996	(3.33)	2898	(96.73)	98	(3.27)
Maternal characteristics
Maternal age (years)
15–19	298	(0.33)	284	(95.30)	14	(4.70)	<0.0001
20–24	4151	(4.61)	4053	(97.64)	98	(2.36)
25–29	28 218	(31.33)	27 732	(98.28)	486	(1.72)
30–34	42 431	(47.11)	41 493	(97.79)	938	(2.21)
35–39	13 176	(14.63)	12 739	(96.68)	437	(3.32)
40+	1798	(2.00)	1686	(93.77)	112	(6.23)
Income level
1Q	8478	(9.41)	8245	(97.25)	233	(2.75)	0.0464
2Q	13 188	(14.64)	12 875	(97.63)	313	(2.37)
3Q	23 596	(26.20)	23 070	(97.77)	526	(2.23)
4Q	29 359	(32.60)	28 711	(97.79)	648	(2.21)
5Q	15 451	(17.15)	15 086	(97.64)	365	(2.36)
Type of insurance
Self-employed insured	26 237	(29.13)	25 556	(97.40)	681	(2.60)	<0.0001
Employee insured	63 551	(70.56)	62 162	(97.81)	1389	(2.19)
Medical aid	284	(0.32)	269	(94.72)	15	(5.28)
Residential area
City	63 425	(70.42)	62 008	(97.77)	1417	(2.23)	0.0130
Rural	26 647	(29.58)	25 979	(97.49)	668	(2.51)
Working status
Work	25 068	(27.83)	24 499	(97.73)	569	(2.27)	0.5771
Not work	65 004	(72.17)	63 488	(97.67)	1516	(2.33)
Prenatal care
Adequate	71 522	(79.41)	69 935	(97.78)	1587	(2.22)	<0.0001
Intermediate	16 730	(18.57)	16 299	(97.42)	431	(2.58)
Inadequate	1820	(2.02)	1753	(96.32)	67	(3.68)
Mode of delivery
Spontaneous vaginal delivery	31 996	(35.52)	31 611	(98.80)	385	(1.20)	<0.0001
Instrumental delivery	24 649	(27.37)	24 228	(98.29)	421	(1.71)
Caesarean section delivery	33 427	(37.11)	32 148	(96.17)	1279	(3.83)
Parity
1 (Nulliparous)	60 109	(66.73)	58 540	(97.39)	1569	(2.61)	<0.0001
2	26 581	(29.51)	26 127	(98.29)	454	(1.71)
3+	3382	(3.75)	3320	(98.17)	62	(1.83)
Twin birth status
Singleton	88 967	(98.77)	87 000	(97.79)	1967	(2.21)	<0.0001
Twin	1105	(1.23)	987	(89.32)	118	(10.68)
Comorbidities during pregnancy
0	89 094	(98.91)	87 218	(97.89)	1876	(2.11)	<0.0001
1+	978	(1.09)	769	(78.63)	209	(21.37)
Hospital characteristics
Type of hospital
Clinic	42 549	(47.24)	41 975	(98.65)	574	(1.35)	<0.0001
Hospital (30 ≤ beds < 100)	15 831	(17.58)	15 636	(98.77)	195	(1.23)
Hospital (100 ≤ beds < 500)	16 468	(18.28)	16 217	(98.48)	251	(1.52)
General hospital(beds 500+)	15 224	(16.90)	14 159	(93.00)	1065	(7.00)
Hospital ownership
Public	394	(0.44)	382	(96.95)	12	(3.05)	0.3336
Private, for-profit	89 678	(99.56)	87 605	(97.69)	2073	(2.31)
Teaching status
Not teaching	85 774	(95.23)	84 203	(98.17)	1571	(1.83)	<0.0001
Teaching	4298	(4.77)	3784	(88.04)	514	(11.96)
Year
2003	8386	(9.31)	8175	(97.48)	211	(2.52)	0.0034
2004	8361	(9.28)	8135	(97.30)	226	(2.70)
2005	8077	(8.97)	7865	(97.38)	212	(2.62)
2006	8112	(9.01)	7932	(97.78)	180	(2.22)
2007	8900	(9.88)	8681	(97.54)	219	(2.46)
2008	8166	(9.07)	8005	(98.03)	161	(1.97)
2009	7429	(8.25)	7251	(97.60)	178	(2.40)
2010	7795	(8.65)	7605	(97.56)	190	(2.44)
2011	8478	(9.41)	8301	(97.91)	177	(2.09)
2012	8623	(9.57)	8437	(97.84)	186	(2.16)
2013	7745	(8.60)	7600	(98.13)	145	(1.87)

Table 2 presents the association between SMM and day/time of delivery unadjusted and adjusted all covariates. In adjusted model, women who delivered on a weekend or holiday and at night on a weekday had significantly higher risks for SMM compared with those who delivered during the daytime on a weekday (weekend: RR 1.26, 95% CI 1.10–1.46, P = 0.0013; weekday night time: RR 1.58, 95% CI 1.30–1.93, P < 0.0001). In addition, we found that the following factors were associated with high risk of SMH compared to reference group: being maternal at younger or older age, having low income level, living in rural area, inadequate prenatal care, caesarean section, nulliparous, twin birth and comorbidities during pregnancy.

Table 2

The association between day/time of delivery and risk of severe maternal morbidity

	Severe maternal morbidity
	Risk ratio	95% CI	P value
Day/time of delivery
Weekday daytime	1.00
Weekday night time	1.26	(1.10–1.46)	0.0013
Weekend or holiday	1.58	(1.30–1.93)	<0.0001
Maternal characteristics
Maternal age (years)
15–19	2.10	(1.29–3.43)	0.0028
20–24	1.37	(1.11–1.69)	0.0031
25–29	1.00
30–34	1.17	(1.05–1.30)	0.0054
35–39	1.37	(1.20–1.56)	<0.0001
40+	1.90	(1.55–2.33)	<0.0001
Income level
1Q	1.31	(1.11–1.54)	0.0013
2Q	1.18	(1.02–1.37)	0.0286
3Q	1.12	(0.98–1.27)	0.0963
4Q	1.07	(0.94–1.21)	0.2966
5Q	1.00
Type of insurance
Self-employed insured	1.09	(0.99–1.19)	0.0793
Employee insured	1.00
Medical aid	1.36	(0.84–2.19)	0.2078
Residential area
City	1.00
Rural	1.17	(1.06–1.28)	0.0010
Working status
Work	1.00
Not work	1.09	(0.99–1.20)	0.0909
Prenatal care
Adequate	1.00
Intermediate	0.99	(0.89–1.11)	0.8759
Inadequate	1.29	(1.02–1.64)	0.0332
Mode of delivery
Spontaneous vaginal delivery	1.00
Instrumental delivery	1.36	(1.19–1.56)	<0.0001
Caesarean section delivery	2.46	(2.19–2.76)	<0.0001
Parity
1 (Nulliparous)	1.36	(1.22–1.51)	<0.0001
2	1.00
3+	1.05	(0.81–1.35)	0.7060
Twin birth status
Singleton	1.00
Twin	1.78	(1.48–2.15)	<0.0001
Comorbidities during pregnancy
0	1.00
1+	5.73	(5.03–6.52)	<0.0001
Hospital characteristics
Type of hospital
Clinic	1.10	(0.93–1.29)	0.2658
Hospital (30≤beds<100)	1.00
Hospital (100≤beds<500)	1.26	(1.05–1.52)	0.0150
General hospital(beds 500+)	3.54	(3.01–4.17)	<0.0001
Hospital ownership
Public	1.00
Private, for-profit	1.75	(1.01–3.05)	0.0478
Teaching status
Not teaching	1.00
Teaching	1.98	(1.77–2.23)	<0.0001
Year
2003	1.00
2004	1.15	(0.96–1.38)	0.1262
2005	1.15	(0.95–1.38)	0.1466
2006	1.10	(0.90–1.34)	0.3439
2007	1.17	(0.97–1.41)	0.1057
2008	0.96	(0.78–1.17)	0.6602
2009	1.18	(0.96–1.44)	0.1168
2010	1.17	(0.95–1.43)	0.1370
2011	0.99	(0.80–1.23)	0.9377
2012	0.99	(0.80–1.23)	0.9559
2013	0.87	(0.70–1.09)	0.2374

Table 3 presents the results of sub-group analysis on most frequent five sub-indicators of SMM by day/time of delivery. Off-hour delivery, both at night time and on a weekend or holiday, were associated with increased high risk of needing a blood transfusion (RR 1.32, 95% CI 1.12–1.56, P = 0.0010 and RR 1.60, 95% CI 1.27–2.03, P < 0.0001, respectively), and delivering on a weekday at night time increased the risk for eclampsia by 1.29 times (RR 1.29, 95% CI 1.11–1.50, P = 0.0008).

Table 3

The association between sub-indicators of severe maternal morbidity and day/time of delivery

	Weekday daytime	Weekday night time			Weekend or holiday
Sub-indicators of SMM	Risk ratio	Risk ratio	95% CI	P value	Risk ratio	95% CI	P value
Overall SMM	1.00	1.26	(1.10–1.46)	0.0013	1.58	(1.30–1.93)	<0.0001
Blood transfusion	1.00	1.32	(1.12–1.56)	0.0010	1.60	(1.27–2.03)	<0.0001
Disseminated intravascular coagulation	1.00	1.01	(0.70–1.44)	0.9743	1.49	(0.96–2.31)	0.0777
Shock	1.00	1.40	(0.84–2.33)	0.1952	1.24	(0.62–2.50)	0.5436
Sepsis	1.00	1.33	(0.71–2.48)	0.3751	1.15	(0.42–3.12)	0.7841
Eclampsia	1.00	1.29	(1.11–1.50)	0.0008	1.14	(0.89–1.47)	0.2982

SMM: severe maternal morbidity.

Adjusted for all covariates.

Figure 1 presents the result of sub-group analysis by provision factors and day/time of delivery. Women who delivered in general hospitals with more than 500 beds, and who delivered weekday night time and weekend were at greater risk of SMM (RR 1.45, 95% CI 1.22–1.73 and RR 1.73, 95% CI 1.35–2.21, respectively).

Figure 1

The relationship between day/time of delivery and severe maternal morbidity by hospital volume and type. *P<0.0.05, ^**P<0.01, ^***P<0.001

Discussion

In this study, we confirmed that off-hour delivery either on a weekday at night time or on a weekend or holiday was associated with SMM. Women who delivered on weekdays at night time, on weekends or on holidays had a significantly higher incidence rate of SMM. In addition, women who delivered at an extremely young or old age, who lived in rural area, who delivered by caesarean section, who were nulliparous, who had multiple births and who had comorbidities during pregnancy had a high risk of SMM. Moreover, we found that a positive relationship existed between off-hour delivery and provision factors for risk of SMM.

The incidence of SMM was 2.31% of total maternity in this study. This was similar to some previous research although other studies had different results. Previous studies reported 2–2.5% of their samples with SMM,^12,29,30 although other studies showed 0.3–1.7% of their study populations with SMM.^13,15,31 The incidence of SMM differed because of differences in the samples, maternal health conditions in the countries or communities and the use of different SMM indicators such as the CDC’s algorithm or WHO’s indicators. This study used the CDC’s algorithm as an SMM indicator, and the observed incidence of SMM was similar to other studies using the same SMM indicators. Additionally, we confirmed the relationship between risk factors and the sub-indicators of SMM. Blood transfusion was the most frequent indicator in almost 60% of all cases of SMM (Supplementary table S1), which is similar to the New York City Department of Health and Mental Hygiene’s report.¹⁴

This study confirmed that off-hour delivery and risk of SMM were significantly associated. In our result, women who delivered on a weekday at night time and on weekends or holidays had a 1.26 and 1.58 times, respectively, higher risk of SMM compared with those who delivered on a weekday in the daytime. In previous studies, Snowden et al. reported that severe maternal and neonatal complications increased on high-volume days and weekends after adjusting for maternal demographics, annual hospital birth volume and teaching hospital status.¹⁹ Palmer et al. showed that puerperal infection was higher for deliveries that occurred on Saturday than on Tuesday.²² In addition, Lyndon et al. observed that night-time birth and SMM had positive association.²³ Moreover, in other medical fields, off-hour hospital admissions have been related to an increased risk of mortality.³²

The aetiology of the off-hour effect remains unclear. In prior research, if the effect was caused by a staff deficiency, a lack of resources, preparedness for emergent situation and less experienced physicians manning the weekend shifts, one would expect poorer quality and safety during all out-of-hours periods during the week.^22,33 Palmer's study provided some evidence to support the theory that one of the contributing factors to the weekend effect might be a failure to meet the recommended levels of consultant presence, as evidenced by a significant association between staffing and perinatal tear rates.²² There might be an explanation of some of the mechanisms underlying the increased risk of SMM on weekdays at night time and weekend deliveries, even though the exact reasons for the weekend effect are not known. However, another possible explanation could be that nursing and physician staffing might include less experienced people, or there are fewer nurses and doctors on off-hour shifts in the hospital. In Lyndon’s research, women who give birth at night may miss the opportunity to recognize and mitigate complications early, which may lead to less preventable serious morbidity.²³ Fatigue and sleep deprivation, cognitive impairment, reduced staffing and fewer resources on night-time shifts may limit the clinical team's capacity to respond to changes in patients’ clinical status.²³ In another study, surgeons often operate with unfamiliar assisting staff during the weekend, which may affect their efficiency and communication.³⁴ Also, increasing number of residents and decreasing level of nurses and physicians on weekend shifts might be related to weekend mortality in non-elective hospitalization.³³

In the sub-group analysis (figure 1), clinics and general hospitals were associated with increased risk of SMM during off-hour deliveries. We cannot definitively determine why there were differences in the risks seen in some facilities at night and in others on the weekends, but these differences might be explained by staffing or limited resources in hospitals at off-hours. In Cram’s study, patients admitted on weekends had a slightly higher risk-adjusted mortality compared with those on weekdays.³⁵ The weekend effect reflected more than physician availability on weekends. It more likely reveals integrated factors that result in reduced quality of care, decreased levels of staffing, reduced availability of certain procedures and an overall reduction in patient supervision when hospital staffing decreases during the weekend.³⁵ Similarly, SMM might occur through delivery failure or failure of early postpartum management caused by a lack of resources and staff during off-hours.

Regarding income level, women with lowest income level had higher risk of SMM compared to those with highest income level. A previous study showed that women of low socioeconomic status (SES) were more likely to have high risk of SMM compared to high-SES women.¹¹ This could be due to the following factors: women in the lowest SES group commonly report poorer experiences of care during pregnancy, have higher risks of prenatal hospital admission and receive less prenatal care or are less likely to meet a midwife or GP for a 6–8-weeks postnatal review.³⁶ As a result, disparities in health-seeking behaviour, access to maternity services and treatment of women by healthcare professionals could contribute to maternal health outcomes.³⁷

In fact, South Korea is undergoing a reduction of obstetricians and labour facilities because of extremely low total fertility rate (i.e. 1.05 in 2017) and low insurance reimbursement for delivery procedures. The risk for SMM during off-hours is particularly of concern for those in vulnerable areas with limited obstetric facilities. In this study, women who lived in rural areas were statistically at higher risk of SMM compared with those who lived in the city area. These phenomena could reflect a disparity in quality and safety of care as well as accessibility; as a result, such disparities might affect maternal health outcome including maternal death.

This study has several limitations. First, for assessment of SMM, we used administrative data (ICD-10) that do not include relevant clinical data on the severity of illness; therefore, we did not define the severity of SMM. Furthermore, we used a published algorithm to identify SMM cases and did not conduct a medical chart review for case ascertainment. Second, we could not adjust for potential confounders such as maternal education level, body mass index and behavioural risk factors (smoking or alcohol drinking), which the data did not contain. Nevertheless, we performed a population-based cohort study and could construct a risk-adjustment that included important confounders available in our linked data set. Third, the data did not contain gestation commencement dates, so the duration of pregnancy could be not calculated precisely. However, because our data contain the exact date of birth, we could estimate the first pregnancy period by calculating the gestation period. Fourth, there is a problem with conversion of ICD-9 procedure codes. The NHIS-NSC used ICD-10 codes that did not include procedure codes; therefore, we converted ICD-9 procedure codes to EDI codes in this dataset. During the conversion process, some cases had no information with an EDI code; therefore, some procedure-based SMM cases might have been less exact.

Nevertheless, this study has some strength. First, to our knowledge, this is the first study of SMM considering quality factors in Korea and the first study of off-hour delivery on SMM. Until recently, there has been little empirical evidence regarding SMM and its association with any quality indicators. This study provides important evidence for use in future maternal health care. Second, it has a population-based design, long-term follow-up and data that were obtained from the NHIS-NSC that are nationally representative. Third, we tried to use objective indicators or databases to adjust for various health care quality factors, particularly, the Kessner Adequacy of Prenatal Care Index and exact delivery day and time could be estimated from NHIS-NSC data. Fourth, we tried to consider various obstetric comorbidities and provision factors to adjust for case mix.

Off-hour delivery was found to be a risk factor for SMM independent of other known sociodemographic, obstetric and provision risk factors. Therefore, policy makers should provide financial and systemic support to allocating adequate human resources and increase the presence of labour facilities in vulnerable areas as well as generally increasing staffing during weekends or the night time to improve the quality of intrapartum and postpartum maternal care. Furthermore, more research on these topics is needed, and future research should examine that whether the staffing and number of physicians, hospital resource and the level of nursing grade affect maternal health outcomes during off-hour periods, and also whether these phenomena affect the regional disparity on maternal health outcome. Such information would provide additional evidence that could help identify individual- and hospital-level policy to improve maternal health outcomes of childbirth.

Acknowledgements

The authors would like to thank the colleagues at the Institute of Health Services Research of Yonsei University, who have provided their advice on intellectual content.

Funding

The authors declare that they received no funding for the completion of this study.

Conflicts of interest: The authors report no conflict of interest.

Key points

• SMM is a more useful indicator for evaluation and improvement of maternal health services.

• Weekend or night-time delivery was related to high risk of SMM, and low income level or living in rural area was also related to SMM.

• Financial and systemic support are needed to allocate adequate human resources and increase the presence of labour facilities in vulnerable areas and generally increasing staffing during weekends or the night time to improve the quality of maternal care.