Maternal Mortality Reduction • Pilot: Months 1-9
Community Health Aides in Bungoma County, Kenya • 2018 – Present
Introduction
In 2018, Agile HMS teamed with two private hospitals in rural Bungoma County to pilot a program focused on improving maternal and neonatal health outcomes in the Kenyan private and public sector by mobilizing, supporting and incentivizing Community Health Aides within the structure of Kenya’s public funded maternal health scheme: Linda Mama. This ongoing pilot seeks to demonstrate the value of employing electronic health records, enterprise workflows, machine learning and continuous system analysis, while utilizing Community Health Aide (CHA) networks to:
Early identify and monitor pregnancies
Detect risk while assuring timely clinical intervention and follow-up
Inform on quality of care outcomes and obstacles to care
This pilot seeks to lay a foundation for the creation of an Early Detection, Management and Research Center (EDMRC) for Maternal and Newborn Disease, designed to support Kenya’s noteworthy initiatives to achieve the WHO 2030 Sustainable Development Goal for Maternal Mortality Rate (MMR) of 140 deaths per 100,000. The envisioned EDMRC addresses the critical requirement for a centralized system of detection and management focused on empowering local providers and community members, who in turn, are better able to achieve state-of-the-art standards of care and patient interaction, and who are equipped with an enterprise-level technology platform that integrates them into a national or international system of management and research. We believe this envisioned ability to quantify and respond to primary causes of clinical and non-clinical risk associated with maternal mortality and injury, is is a critical component of Kenya’s ability to achieving the 8.26% average annual rate of reduction (ARR) in MMR, necessary to achieve the WHO 2030 Sustainable Development Goal for MMR.
During the first nine months of the pilot, 445 mother were enrolled and 253 births were reported, with no maternal or new born deaths.
Method
As stated above, our goal is to utilize CHA networks to better inform on information required for enterprise level manage of micro-systems, while leveraging recent inclusion of private provider reimbursement for Kenya’s Linda Mama program, to create a private clinic revenue-model which justified clinic funding of CHA networks.
We observe that Linda Mama reimbursement, which total approximately $100 per pregnancy (assuming four antenatal and three postnatal checkups, and normal birth) to properly managed private clinics, are both profitable and of value to the mother, and it is advantageous for clinics to increase their maternal case load. To this system cost, we envisioned and have now implemented a payment through the private hospital, to individual CHAs for the critical and collaborative role they fulfill in identifying and managing pregnancies. This payment, coupled with training and becoming a valued member of the formal medical delivery ecosystems, has provided several positive results, namely, greater uptake of antenatal checkups (ANC) to a population of mothers with limited or no historical access to ANC. The fee paid to CHAs is similar to the traditional cost of rural home delivery in Kenya, and totals approximately 5.27% of the normal clinical reimbursement by Linda Mama.
Caregivers and facility administrators from Bungoma have provided valuable insight and guidance on our pilot program design and implementation, relevant to the local population and existing health system. As far back as October 2016, we – with MoH oversight – facilitated workshops in Bungoma and met with over 50 healthcare worker from 7 facilities specializing in maternity care. The workshops included MoH officials in Bungoma, along with nursing officers, midwives, hospital facilitators and community health workers, and resulted in program design (enterprise platform, data collection and training of program participants) addressing three specific requirements that severely impact local and national maternal and new born objectives:
The need for mothers to receive early, quality clinical care
The need to manage and distribute demand among facilities
The effective and efficient distribution of maternal and neonatal resources
Maternal Clinical Data Analysis: Months 1-9
Here we present our analysis of the trends in healthcare access and outcomes from the maternal clinical data. Overall, during the period of observation, across the two hospitals, a total of 445 women were seen for pregnancy related services. The mean age for the entire sample was 25.73 years (SD = 6.21 years). For these women, the total number of births recorded at the hospitals totaled 253, where the mean age for the mothers who gave birth was 25.38 years (SD = 5.93 years). There were no reported maternal or infant deaths. Of the 253 births, 12 mothers were reported as having been referred to other institutions for complications, 170 were reported as having not been referred, and no referral status was recorded for 71. Similarly, 11 infants were reported as being referred to other institutions for complications, 171 were reported as having not been referred, and no referral status was recorded for 71. Based on estimated due dates, of the 253 births, 25 were recorded as have been less than 37 weeks gestational age at birth, and 20 were recorded as having been 43 weeks or greater gestational age at birth. Gestational age at birth was unreported for 18 infants. At the time of birth, 66 of the 253 mothers who gave birth had 4 or more recorded ANC visits. This rate, 26%, is similar to that reported as part of the Ministry of Health’s Confidential Enquiry into Maternal Deaths in Kenya which noted, of the women in that study who attended an ANC visit only 1 in 5 had at least 4 visits (Republic of Kenya, Ministry of Health, 2017, Saving Mothers Lives, Confidential Enquiry into Maternal Deaths in Kenya.). However, this rate falls below the rate of 51% reported by UNICEF for rural Kenya as of 2014 (UNICEF, 2018, Maternal and Newborn Health Coverage Database).
Time-Based Trends. The table below shows the frequency counts related to the total volume of maternity-related activity at the hospitals by month during the complete study period. With respect to births, there were an additional 29 recorded that were either missing full birth date information or that occurred prior to September, but the mothers were seen for follow up visits in September or later. Similarly, for new enrollments, an additional 33 occurred prior to September but the women were seen for additional appointments in September or later.
Overall, across the study period, new enrollments peaked in December and then tended to fall off over time with some variability, and similarly, maternity visits peaked in November and then tended to fall off over time with some variability and a sharp rebound in May. Births also tended to peak in January and then fall off over time with some variability. The relationship between these trends and the measures of online media activity has already been discussed. However, it is also important to note a few factors that likely impacted rates of activity as well. First, the hospitals reported that due to disruptions within Kenya, no claims were paid during the study period until March, and then these payments started to catch up but were not complete. Likewise, maternity specific claims to NHIF were never paid after December 2018 through the duration of the study period. Hence, these disruptions caused staffing problems and pharmacy limitations at the hospitals that were exacerbated over time, although staffing problems reportedly started to return to normal in May. Likewise, these payment disruptions also affected the activities of Community Health Aides who may have otherwise been able to encourage additional pregnant women to seek care. In April in particular the hospitals noted staffing shortages, but at the same time also noted a generally slow flow of mothers seeking maternity related services during that same period.
In addition, trends with respect to the recency of visits also varied over time. The table below illustrates these trends as percentages for women who were pregnant during each reporting month, starting in November when data collection had been in place for a few months. Note that the distribution shifts later over the first few months as the data started to “age” following the start of the study and hence the first few months of recording these variables can be seen as a transient. Nevertheless, it is clear from these data that timeliness tended to bottom out in April, and then rebound in May, consistent with the maternity visits and births trends noted above.
In addition, with respect to attendance at recommended visits, there was also increasing trend toward being late or not attending at all. In particular, the table below shows that for the women who were pregnant, there was a trend toward a small drift in the percent late or not attending the ANC visits as the study period progressed. However, trends also indicated that there was some month to month variability, and in addition, some suggestion of improvement by May for ANC 3 and 4.
Overall, these time based trends generally tell a similar story over time, where overall maternity related activity peaked in the third to fifth month of the nine month study period (November to January), and then fell off with some suggestion of a rebound in the final month of the study (May). Similarly, recency of visits and adherence to the schedule of recommend visits also tended to worsen during the study period, once again with some suggestion of improvements in May.
Differences in Key Groups. Beyond these time-based trends, although not the primary focus of the work reported here, we also sought to better understand particular groups of interest that were potentially at risk in order to help inform future efforts. The first group were mothers who gave birth at a hospital with no record of prior ANC visits (0 ANC Births), as there was no indication of these women receiving prenatal care at one of the participating hospitals prior to birth. Second, we looked at women who were referred out for additional care following birth due to complications. Finally, we also looked at a group of women who attended ANC visits but who never delivered in the participating hospitals. These women were of interest because although they had some prenatal care, they likely gave birth outside of the setting of the participating hospitals. Across these various groups, there may be opportunities for future targeted engagements, and accordingly, here we explore the extent to which we could determine trends based on a limited set of available data.
The following table shows data for mothers with recorded births. The first column shows descriptive statistics for all of the birth mothers in the sample, the second column shows data for women with 0 reported ANC visits, and the final column shows data for the women who were referred out for follow up following birth. Trends suggested that relative to all birth mothers, the 0 ANC birth mothers tended to be slightly older and to be more likely to be married. In contrast, relative to all birth mothers, data suggested that the referred mothers tended to have additional ANC visits and tended to have a history of fewer pregnancies.
Table 5 shows similar data but for the women in the study who had not yet been recorded as having given birth as of the end of the study. The first column shows data for all of the pregnant women, and the second column show data for the 45+ group. This later group included anyone with a gestational age of 45 weeks or more with no record of a birth, and could have included those who delivered outside a participating hospital or for which a pregnancy terminated for other reasons. There are no clear trends in terms of differences for this group with respect to all pregnant women. However, relative to the mothers who had given birth (all birth mothers), there is a slight trend toward being older. In addition, a notable finding is the size of the 45+ group in that it had an N of 77, indicating that of the 445 total mothers seen during the study period, 17% had pregnancies that ended outside of the participating hospitals. Also notable was that for these 77 mothers, 43 (56 %) had only one ANC, suggesting minimal contact with the hospitals. Yet, of these 77 mothers, another 34 (44%) had more than one ANC visit and 16 (21%) even had the 4+ suggested visits, suggesting extensive contact with the hospitals despite eventual delivery outside the hospital setting.
Overall, although only suggestive, these data indicate that older women may be more likely to not attend ANC visits and not deliver inside a hospital setting. In addition, married women may be more likely to not attend ANC visits. In terms of mothers who did not deliver in the hospitals, the data suggest differences where some mothers may seek minimal care from a hospital and may then instead rely on other care. However, other women may seek care, but be unable to deliver in a hospital setting. Anecdotal reports from the hospitals indicate that some of these women may have had issues with transportation. Finally, for women who were referred, data suggest that they may be less experienced in terms of number of pregnancies, and importantly, the trend toward additional ANC visits indicates that these women tended to get additional prenatal care. Collectively, therefore, these data suggest a set of differences that impacted care and that strategies for intervention, including communication regarding available resources, likely needs to be differentiated to be maximally effective. Additional research that focuses on exploring the differences noted here along with additional demographic factors is suggested to verify and extend these findings.
Discussion
While our limited data set of only two CHA networks supporting two private hospitals addresses: 1) the need for women to receive early, quality clinical care, 2) the need to manage and distribute demand among facilities, and 3) the effective and efficient distribution of maternal and neonatal resources, it is not yet adequate to draw broad conclusions. However, two observations readily present themselves related to the operational aspects of our approach.
The CHA networks appear to be highly influenced by payment, as we observed a rapid spike of ANCs following full implementation of the payment program. This was followed by a sharp, though not to pre-payment levels, drop in clinical visits when CHA payments were delayed due to system delays in reimbursement, with observed evidence that CHAs sought out public hospitals that also offered some form of compensation for CHA, even if at significantly greater distances of travel for the patient. We have recently learned from our CHA networks, the NGO funding source for CHA payment in the Bungoma public sector has ceased operations, and our CHA networks plan to return to our clinics.
Despite significant challenges with reimbursement from Linda Mama, the program was able to fairly rapidly achieve 2.14 ANCs prior to delivery with 44% of mothers receiving 3 or more ANCs prior to delivery, within a rural population that has no recorded history of ANC accessibility. The continued relative high rate of clinical visits (record updated) to births suggests local mothers increasingly value access to ANCs.
Acknowledgments
We would like to offer special thanks to the Community Health Aides and the clinical staff of our teaming partner, for their continued commitment to community engagement, collection of critical data and tireless efforts to identify maternal and new-born risk. We are grateful to join you in your efforts to create a a healthy and thriving community.
Funding
The clinical portion of this pilot is funded by National Hospital Insurance Fund, Republic of Kenya as part of the Linda Mama initiative; pilot management, to include training and compensation for Community Health Aides, is funded by Sophia Health Systems, through its subsidiary Agile HMS