Standardizing the measurement of maternal morbidity: Pilot study results

Abstract Objective To field test a standardized instrument to measure nonsevere morbidity among antenatal and postpartum women. Methods A cross‐sectional study was conducted in Jamaica, Kenya, and Malawi (2015–2016). Women presenting for antenatal care (ANC) or postpartum care (PPC) were recruited if they were at least 28 weeks into pregnancy or 6 weeks after delivery. They were interviewed and examined by a doctor, midwife, or nurse. Data were collected and securely stored electronically on a WHO server. Diagnosed conditions were coded and summarized using ICD‐MM. Results A total of 1490 women (750 ANC; 740 PPC) averaging 26 years of age participated. Most women (61.6% ANC, 79.1% PPC) were healthy (no diagnosed medical or obstetric conditions). Among ANC women with clinical diagnoses, 18.3% had direct (obstetric) conditions and 18.0% indirect (medical) problems. Prevalences among PPC women were lower (12.7% and 8.6%, respectively). When screening for factors in the expanded morbidity definition, 12.8% (ANC) and 11.0% (PPC) self‐reported exposure to violence. Conclusion Nonsevere conditions are distinct from the leading causes of maternal death and may vary across pregnancy and the puerperium. This effort to identify and measure nonsevere morbidity promotes a comprehensive understanding of morbidity, incorporating maternal self‐reporting of exposure to violence, and mental health. Further validation is needed.

and the Global Strategy to End Preventable Maternal Mortality (EPMM), 1,2 improving the measurement of maternal health will be key. Building on the success of the WHO in defining and measuring maternal near-miss events, 3 which also established parameters for quality of care for severe maternal complications/morbidities, action is now focused on standardizing and measuring non-life-threatening maternal morbidity.
As maternal mortality trends downward, measuring morbidity will be critical to monitoring the quality of maternal health care. Based on previous efforts of the WHO Department of Reproductive Health and Research (RHR) to standardize the measurement and reporting of maternal mortality and severe maternal morbidity, efforts are expanding to also address measurement of non-life-threatening (nonsevere) maternal morbidity, especially given that the often-cited maternal morbidity estimate of around 20-30 morbidities for every maternal death is "not based on standard, well documented, and transparent methodologies". 4 RHR implemented a 5-year project to address the lack of a scientific basis for defining, estimating, and monitoring the magnitude of maternal morbidity. It was envisioned that in addition to standardizing what is called maternal morbidity and how it is measured, doing so would not only assist program managers and policy makers to better monitor maternal morbidity, but it would also bring attention and resources to enhancing care for pregnant and postpartum women. The present paper describes the initial performance of this new set of standardized maternal morbidity measurement tools, and represents the final step in a larger initiative (described in the methods section) that seeks to advocate for and improve women's health.
The pilot study sought to field test a comprehensive instrument to measure nonsevere morbidity among women in antenatal care (ANC) and postpartum care (PPC). The paper presents study findings and insights into its future use. Subobjectives included: 1. Describing the sociodemographic characteristics of the women recruited, by country.

Examining the contributory factors and clinical indicators identified
and their relationship to obstetric and medical diagnoses among pregnant and postpartum women.

3.
Exploring the feasibility and challenges associated with administering the instrument (for research purposes only).

| MATERIALS AND METHODS
This study is the culmination of a 5-year initiative led by the Maternal Morbidity Working Group (MMWG) convened by RHR, the details of which have been published elsewhere. [4][5][6] In brief, the group, composed of technical experts in maternal and women's health, began by defining non-life-threatening (hereafter referred to as nonsevere) maternal morbidities as: "Any health condition attributed to and/or complicating pregnancy and childbirth that has a negative impact on the woman's wellbeing and/or functioning". 4 To operationalize this definition, a maternal morbidity matrix was developed. 4  For this pilot, the aforementioned matrix was translated into two comprehensive questionnaires: the ANC and PPC maternal morbidity measurement instruments. The tools sought a holistic view of maternal health, guided by women's perspectives. Previously validated scales were incorporated, where available, to measure mental health (General Anxiety Disorder, 7 item (GAD-7), 9 and the Personal Health Questionnaire, 9 item (PHQ-9) 10 ); health-related functioning, or the ability to carry out daily tasks and social responsibilities (using the WHO Disability Assessment Schedule (WHODAS 2.0) 12-item version 11 ); sexual satisfaction 12 ; substance use/abuse 13 ; and exposure to violence. 14 Both instruments were hour-long questionnaires consisting of a patient interview, physical examination, and record review. 15 The interview documented socioeconomic status, medical and obstetric history, and clinical symptoms. The physical examination, conducted by a healthcare professional, evaluated clinical signs. The record review extracted information on selected laboratory tests and results.
To field test the instruments, a cross-sectional study was conducted in three countries (Jamaica, Kenya, and Malawi) in late 2015 to early 2016, over 3-month periods in each country. All sites (n=13) were public facilities: nine in Jamaica (6 health centers and 3 referral hospitals), three in Kenya (2 district and 1 referral hospital), and one in Malawi (referral hospital). Efforts were made to include a range of facilities (primary, secondary, and tertiary referral facilities; in urban and rural settings) from different subnational areas, although the choice also depended on availability of the relevant resources to ensure highquality data collection. Ethical approval was obtained from the WHO Ethical Review Committee and relevant entities in each country. To describe the different types of morbidity, and enable stratification by country setting and time of administration, a sample size of 500 women per country (250 each, ANC and PPC) was calculated to be adequate. Without pooling data across sites or populations, we estimated a 6% margin of error.
Participants were conveniently selected from among women presenting for routine ANC or PPC services. Inclusion criteria were that ANC women were at least 28 weeks pregnant, and PPC women were 6-12 weeks after delivery. 15 Women whose pregnancies ended in abortion or miscarriage were excluded, while those who experienced stillbirths were included. Women provided written informed consent before being interviewed.
Health professionals (nurses, midwives, or doctors, depending on the site) participated in a 2-day training session to administer the ANC and PPC instruments. They were informed about the study objectives, introduced to the instruments, learned how to use the digital tablets, and gained practical experience before final selection. A manual detailing standard operating procedures for each question was developed and modified for each country. In Jamaica and Kenya, the interviewers were facility staff recruited for this project. In Malawi, ANC nurse midwives conducted the physical exam, while recent medical graduates were hired specifically to conduct the interview portion of the tool.
The instruments were developed in digital format and translated into local languages at each site. Data were gathered on tablets using OpenDataKit (University of Washington, USA) open-source software. The digital tool limited the prevalence of missing values as interviewers could not advance until an answer was inputted. For women presenting with a condition that warranted referral (mental health score, report of violence, need for further testing, or for procedures for a certain condition), interviewers were provided with referral information on their tablets (such as the department to contact).
Data were stored and managed on a secure, password-protected, cloud- To demonstrate the utility of the tool for research and future health-service intervention, multivariate (logistic regression) analyses were undertaken. Covariate variables were examined by the following groupings.  P<0.001) were recruited from referral sites to ensure that the instrument could be tested on women with a morbid condition in pregnancy and the puerperium.

| RESULTS
In expanding the definition of morbidity, selected contributory factors were explored. We examined exposure to violence by asking women whether they had been "afraid of your current/most recent husband or partner or anyone else," or whether "since pregnancy/ delivery, was there ever a time when you were pushed, slapped, hit, kicked or beaten by (any of) your husband/partner(s) or anyone else?" If women replied affirmatively to either question, they were asked three additional violence-related questions. Thirteen percent (12.8%; n=96, P=0.018) of ANC and 11.0% (n=81, P<0.001) of PPC women reported being afraid of, or having experienced some form of physical violence from, their current partner or someone else, with rates varying by site from 7% to 17%. We also explored substance use through selfreporting (3%-4% overall; Table 3) and, given the global prevalence of obesity, women's heights and weights were documented (Table 4).
Most ANC (61.6%, P=0.018) and PPC (79.1%, P=0.027) women were healthy (i.e. they were not diagnosed with any conditions by healthcare providers, and did not self-report mental health problems). Twenty-seven percent (27.3%, n=205) of ANC and 15.1% (n=112) of the PPC population were diagnosed with one morbidity (including screening positive for depression or anxiety), and 11.1% (n=83) of ANC and 5.7% (n=43) of PPC women had two or more conditions ( with depression in the ANC population, and two women in the PPC population (no women were diagnosed with anxiety by a healthcare provider). Conversely, the GAD and PHQ found that 6.3% (n=47) of ANC and 2.2% (n=16) of PPC women had one or both signs of significant distress ( Table 5). The psychiatric conditions category therefore incorporated diagnosis by a healthcare provider, or scoring 10 or above on the GAD-7 or PHQ-9 questionnaires.
We conducted univariate and logistical analyses for ANC and PPC  Similarly, our clinical examination, and self-reporting of depression and anxiety (using GAD-7 and PHQ-9 for screening), documented a low prevalence. In many settings, mental health concerns are highly stigmatized. Cases may be missed as patients may somaticize these concerns through vague symptoms such as fatigue. 26 A Ghanaian study utilizing the PHQ in a postnatal population found a rate (3.5%; 95% CI, 3.2-3.7) that mirrored the one in our Jamaican population (3.5%; 95% CI, 1.3-5.8). 27 Many social and economic factors have been linked to antenatal and postnatal depression, including experiencing a first pregnancy, 28 being unmarried, 29 exposure to intimate partner violence, 30 and lack of partner/baby father support, 31 or of family cohesion. 28 Identifying and treating these mothers is critical not only to their health but also to the survival and development of their infants. 32 The documented rates require refocusing ANC and PPC services to ensure that these women receive appropriate care.
Limitations of the pilot study included a nonrepresentative sample population, especially in Kenya and Malawi, which was drawn from large urban hospitals. Given this relatively healthy population of women, we had to aggregate many diagnoses, as numbers were too small to separate specific diagnoses. The instrument does not include questions addressing the baby's signs and symptoms, or complications of labor and delivery, which may also help identify morbidity in mothers. Additionally, we were unable to follow the same women over time to measure the temporality of conditions, or distinguish between conditions that developed during pregnancy and prior to pregnancy (i.e. women diagnosed with diabetes mellitus were aggregated into the gestational diabetes mellitus category, as most are first tested and diagnosed during pregnancy). While the current instrument's focus on women's health-related functioning is important, it lacks a wellbeing component to illuminate how women interpret the experience of pregnancy, and how they feel about their health.
The pilot study presents a novel approach to measuring nonsevere maternal morbidity, and furthers the work of WHO and the MMWG to standardize the definition, identification, and measurement process. This holistic approach to assessing maternal morbidity will provide a basis for advocacy for women's health and rights in the broader context. 33 Further research is needed to validate the instrument, and to ensure that the data collected can be used to assess and improve maternal care, especially postpartum visits and continuing health care for women. This is consistent with the EPMM strategy to "address all causes of maternal mortality, reproductive and maternal morbidities and related disabilities" 2 and aligns with the progress needed to

CONFLICTS OF INTEREST
The authors declare no conflicts of interest.