Effectiveness of an Electronic Partogram: A Mixed-Method, Quasi-Experimental Study Among Skilled Birth Attendants in Kenya

BACKGROUND: Timely identification and management of intrapartum complications could significantly reduce maternal deaths, intrapartum stillbirths, and newborn deaths due to hypoxia. The World Health Organization (WHO) identifies monitoring of labor using the paper partograph as a high-priority intervention for identifying abnormities in labor and fetal well-being. This article describes a mixed-method, quasi-experimental study to assess the effectiveness of an Android tablet-based electronic, labor clinical decision-support application (ePartogram) in limited-resource settings. METHODS: The study, conducted in Kenya from October 2016 to May 2017, allocated 12 hospitals and health centers to an intervention (ePartogram) or comparison (paper partograph) group. Skilled birth attendants (SBAs) in both groups received a 2-day refresher training in labor management and partograph use. The intervention group received an additional 1-day orientation on use and care of the Android-based ePartogram app. All outcomes except one compare post-ePartogram intervention versus paper partograph controls. The exception is outcome of early perinatal mortality pre- and post-ePartogram introduction in intervention sites compared to control sites. We used log binomial regression to analyze the primary outcome of the study, suboptimal fetal outcomes. We also analyzed for secondary outcomes (SBAs performing recommended actions), and conducted in-depth interviews with facility in-charges and SBAs to ascertain acceptability and adoptability of the ePartogram. RESULTS: We compared data from 842 clients in active labor using ePartograms with data from 1,042 clients monitored using a paper partograph. SBAs using ePartograms were more likely than those using paper partographs to take action to maintain normal labor, such as ambulation, feeding, and fluid intake, and to address abnormal measurements of fetal well-being (14.7% versus 5.3%, adjusted relative risk=4.00, 95% confidence interval [CI]=1.95-8.19). Use of the ePartogram was associated with a 56% (95% CI=27%-73%) lower likelihood of suboptimal fetal outcomes than the paper partograph. Users of the ePartogram were more likely to be compliant with routine labor observations. SBAs stated that the technology was easy to use but raised concerns about its use at high-volume sites. Further research is needed to evaluate costs and benefit and to incorporate recent WHO guidance on labor management. CONCLUSION: ePartogram use was associated with improvements in adherence to recommendations for routine labor care and a reduction in adverse fetal outcomes, with providers reporting adoptability without undue effort. Continued development of the ePartogram, including incorporating new clinical rules from the 2018 WHO recommendations on intrapartum care, will improve labor monitoring and quality care at all health system levels.

This was a mixed-method, quasi-experimental evaluation of labor management outcomes. Our intervention included a 2-day refresher training for SBAs and supervisors in control and intervention sites and introduction and training on use of the ePartogram in intervention sites. All outcomes except one were during the intervention period, comparing the ePartogram intervention. The exception was outcome of early perinatal mortality before and during ePartogram introduction in intervention sites compared to control sites. The study was conducted from October 2016 to May 2017 in 12 health facilities serving 2 counties, Kisumu in western Kenya and Meru in eastern Kenya. The 2 tertiary care facilities, 1 per county, were allocated to receive the intervention. The other facilities were selected after matching on characteristics like birth volume, staffing level, facility type (public or private), and provision of basic and/or comprehensive emergency obstetric and newborn care (BEmONC and CEmONC, respectively). We allocated 2 large CEmONC facilities to the ePartogram group. The remaining 10 facilities had similar overall delivery rates and were randomly allocated to the intervention or comparison group. Thus, the ePartogram group had 2 large CEmONC and 4 BEmONC facilities, and the paper partograph group had 4 small CEmONC and 2 BEmONC facilities. The Kenyan public health care structure has large central tertiary hospitals, usually 1 per county, and smaller district hospitals and health centers. The larger hospitals usually have a specialist obstetrician-gynecologist and many midwives, and health centers often only have 1 or 2 midwives. Often care is given by providers who do not fit the WHO definition of SBAs. Public services are complemented by faith-based institutions and private maternities. Kenyan health facilities use the WHO modified partograph with alert and action lines; a partograph is started once the woman’s cervix is 4 cm dilated. SBAs are supposed to fill out partographs, but many Kenyan health facilities are poorly staffed, and sometimes partographs are completed by nurses, students, and other non-SBAS who are deployed in labor wards. The ePartogram is an Android tablet-based application developed using human-centered design between 2011 and 2017 to address many challenges of monitoring labor with the paper partograph. The ePartogram is a clinical decision-support tool with algorithms and clinical rules that are based on WHO guidance for managing normal and complicated labor. The app provides auditory reminders to prompt providers to take measurements when due and provides visual and auditory alerts when clinical rules are triggered that predict complications (via low-level alerts) or detect complications (via high-level alerts) (Figure 1). The app allows for increased data entry efficiency by automatically graphing data and storing all client files within the application. In addition, retrospective entry of data is prevented. The decision-supporting software is based on 77 clinical rules based on measures of fetal and maternal well-being, progress of labor, and expected trends as labor progresses. ePartogram Dashboard Screenshot Dashboard shows patients under care of a provider, alerts, and overdue assessments. The ePartogram app was developed to address many challenges of monitoring labor with the paper partograph and improve decision making. The ePartogram app can handle multiple patients who a provider may be taking care of. A dashboard displays all the patients in a provider’s care and highlights those that need immediate attention. In intervention sites, supervisors had access to digitally transmitted ePartogram data anytime on a tablet device if they chose to review and act on them. In control sites, supervisors had access to paper partographs when they visited or were called to the labor ward, which was the supervision norm. Remote supervision off-site with the ePartogram was possible, but we did not activate this function during the study. Data were stored securely in the cloud and were available in real time. The app also prompted the provider to record actions and interventions they took. The app can be tailored to the needs of the country by adding or removing clinical rules. The technical details of the ePartogram are described elsewhere.22 The usability and functionality of the ePartogram underwent field validation exercises in Kenya and Tanzania and a feasibility study in Zanzibar.22 Qualitative feedback from the Zanzibar study informed additional functional enhancements to the ePartogram, including the ability to print a paper record of the ePartogram, inclusion of a dashboard for supervisors and managers, the ability to digitally transfer the full labor record to referral sites, and use of more intuitive icons on the app. Study coinvestigators trained 9 master trainers during a 2-day workshop. This training-of-trainers workshop included interactive presentations and case studies, and it facilitated practice using both the paper partograph and ePartogram. Clinicians meeting the WHO definition of SBA (that is, a doctor, nurse, or midwife “trained to proficiency in the skills needed to manage normal [uncomplicated] pregnancies, childbirth and the immediate postnatal period, and in the identification, management and referral of complications in women and newborns”23) and caring for women in active labor at study sites were recruited. We did not include midwifery students or any unqualified birth attendant even though they often provide labor care in these facilities. All SBAs in a selected facility were recruited to the same study group and gave individual written consent to participate. All participating SBAs and supervisors completed a 2-day labor management update conducted by master trainers and coinvestigators. This training included case studies on using the paper partograph, decision making, and managing normal labor and common labor complications according to WHO and Kenya Ministry of Health guidelines. Additional topics included respectful maternity care; recognition and management of fetal, maternal, and labor progression abnormalities; fetal distress; pre-eclampsia/eclampsia; and fever. Trainees were also refreshed on when supervisors should be contacted in response to labor abnormalities identified, the standard operating procedures for handing over patients as shifts changed, and how to fill out facility birth and outcome registers. We had an additional layer of scrutiny for the ePartogram group only, although this did not involve helping with clinical issues and dealt only with fixing the occasional technology-related issues, such as failure of supervisor tablets to sync with SBA tablets. All participating skilled birth attendants and supervisors completed a 2-day labor management update. We used guided case studies to instruct SBAs on how to interpret clinical information displayed on the partograph. The content and approach for the 2-day labor management training was identical for both study arms. The ePartogram provided prompts to record interventions and actions that the SBA takes, and because this was not available on the paper partograph, we included an action-taken sticker on all paper partographs so that actions could be recorded (Figure 2). Completed Paper Partograph With Stamp Tool to Record Interventions Intervention group participants were also trained in the use of the ePartogram application, maintenance of the tablets, and handover protocols. During the day-long ePartogram training, we provided step-by-step didactic training on the application. We also trained the participants on the standard operating procedures for ePartogram use, storage, and cleaning and on how to print from the ePartogram. We excluded SBAs who did not pass the written assessment during the update training, did not pass the written or skills-based assessments, or did not complete the training. Overall, 72 SBAs were trained in the intervention group, of whom 69 passed the assessments and consented to participate the study. Of the 42 SBAs trained in the comparison group, 1 withdrew from the study. At all study sites, women in active labor received care according to global and national standards. SBAs in intervention facilities used the ePartogram to document measurements; back-up paper partographs were available at each site. In the comparison facilities, SBAs used paper partographs to monitor women in labor. Client names and identifiers from the ePartograms and paper partographs were removed before partographs were scanned. We only used the partographs and labor records filled out by trained, consenting providers for this study. Study outcomes relate to compliance with globally recommended labor-monitoring practices and recording the measurements obtained on the paper partograph and ePartogram; actions taken to maintain normal labor; detection of and actions taken to address deviations from normal labor and complications; and client outcomes. Table 1 describes the indicators used as outcome measures. Outcome Measures Used in the Mixed-Method Quasi-Experimental Study of the Effectiveness an Electronic Partogram Abbreviations: ePartogram, electronic partogram; SBA, skilled birth attendant. When the study was originally conceived, the sample size was calculated at 2,600 based on detecting differences in outcomes among laboring women whose parameters placed them to the right of the alert line (denoting abnormal progress). As the action line has in recent times been increasingly called into question, we abandoned a sample size based on that parameter. In addition, recruitment was disrupted by a health worker strike, resulting in a significantly larger proportion of patients arriving late in labor. Hence, we have presented a power analysis of the final effective sample size. Overall, 1,884 paper partographs and ePartograms were collected. Analysis was done on a total sample of 1,884 and on subsets of this sample where some variables were missing. For some outcomes, we restricted the study analysis to those with 2 or more entries, giving a sample of 1,609. For fetal outcomes, after accounting for missing data, we analyzed a final sample size of 1,405. We performed a post hoc calculation using a type I error of 0.05 and power of 80% to assess the detectable difference. The primary outcome was a composite fetal outcome with a control group prevalence of 18% and an intraclass correlation coefficient of 0.11. The sample size would have been sufficient to detect a 16% difference between the intervention and comparison groups. A trained health records information officer at each facility de-identified and scanned each completed partograph into the study database for cleaning and analysis. All paper partographs had an additional data collection tool stamped on them (Figure 2) or integrated into the electronic app (ePartogram), which asked whether the SBA instructed the client to walk/ambulate, eat food, or drink water during labor; gave the client fluid; encouraged the presence of a birth companion; administered oxygen; changed the client’s position; checked for bleeding; consulted a specialist; referred the client to another site; or performed any other clinical action. Data were also collected about the number of births and maternal and newborn outcomes reported monthly in the maternity ward register. Once research assistants and/or study staff members received these files, they checked them to ensure they were legible, scanned correctly, and titled correctly and then saved the files in a master folder. Research assistants reviewed these scanned partographs and extracted data into the clinical and translational research software, Research Electronic Data Capture (REDCap).24 Research assistants were not blinded to use of the ePartogram/partograph. Each research assistant’s data extraction was validated by a sample of charts that were double entered by another assistant as part of the extraction training process. All data were abstracted by the researchers using REDCap. Data were analyzed in Stata version 14.2.25 For this study, we assumed that missing data in paper partograph and ePartogram measurements were not missing at random; hence, we did not impute them. All estimates presented were based on complete case analysis. We used frequencies and percentages for categorical variables and median to present descriptive statistics for client and facility characteristics across the study groups. We tested hypotheses using nonparametric methods with Fisher exact tests for categorical variables and the test of medians for continuous variables. To assess the intervention’s overall impact, we created a log binomial model with fetal outcome as the dependent variable and ePartogram use as the independent (intervention) variable. We used facility and individual level variables as control variables as part of multivariable regression model to account for the baseline differences between the intervention and comparison groups. The control variables included the type of facility (CEmONC versus BEmONC), affiliation (public versus faith-based), and number of providers at the facility. We also controlled for a nationwide health worker strike that disrupted health services at study sites between December 2016 and February 2017. At the individual level, we used parity (primipara versus multipara) and cervical dilation at admission (≤5 cm versus >5 cm). The adjusted relative risks are presented as the effect size of the ePartogram intervention along with their 95% confidence intervals (CIs). Because laboring women may have been attended to by multiple providers from the facility, we accounted for clustering only at the level of the facilities using Huber-White sandwich estimators. To estimate the intervention’s effect, we compared data obtained from facility registers from 6 months before the study (May–October 2016) and after the implementation of the study (May 2017). All SBAs and supervisors were updated on how to correctly fill out facility registers during the training. In addition, register data were extracted and collected weekly from study-trained health records officers. No other effort was made to ensure completeness or validity of these data. The data points included the monthly aggregates for total births, number of stillbirths, and number of newborn deaths in the first 24 hours for each facility. The final analysis included register data from the 12 facilities on births over 12 months (2 facilities had 0 births in a month). The outcome variable was early perinatal mortality, defined as early neonatal (<24 hours) death and fresh stillbirth over a denominator of the sum of live births and fresh stillbirths. We analyzed for difference between intervention and comparison groups before and after the study by fitting a population-averaged generalized linear model using generalized estimating equations with a Poisson distribution, log link, and exchangeable correlation structure. The models were adjusted using facility-level clustering and adjusts for facility variables of affiliation (public versus private), capacity to address emergencies (CEmONC versus BEmONC), presence of a medical or clinical officer (yes/no), percentage of providers trained in labor management by study staff (≥75% or <75%), median years of provider experience, and whether a health worker strike was occurring in the month in public facilities (yes/no). Incident rate ratios (IRRs) with 95% CIs for before and after rates in each study group are presented along with the difference-in-difference estimator (ratio of IRRs). Four facility in-charges and 28 SBAs from both arms participated in in-depth interviews lasting 30 to 35 minutes. Interviews were conducted in English, recorded, and transcribed. Data collectors were implementing agency staff trained in qualitative research data collection procedures using a data collector training guide26 and study ethics. We coded the qualitative data in Atlas.ti version 7 software,27 and created a codebook based on field guide topics and themes that emerged from the interviews. Interview data saturation was deemed to have been reached, as no new themes emerged in the final interview and all themes were mentioned to some extent in all interviews. The analysis followed the framework analysis process recommended by Ritchie et al.28 We identified themes and subthemes to align with research questions; we describe these later using illustrative quotations. The study was approved by the Johns Hopkins Bloomberg School of Public Health Institutional Review Board (JHSPH IRB #6958) and the Kenya Medical Research Institute (KEMRI protocol #530). SBAs gave informed written consent before participation. Women cared for using the ePartogram did not have a paper partograph plotted, but on completion of labor, a printout was made for the record and for de-identification before scanning to the study database. The JHSPH and KEMRI Institutional Review Boards required that providers participating in the study give consent but did not require that women cared for during labor provide consent.

N/A

• Dima Health