Gaussian process emulation to improve efficiency of computationally intensive multidisease models: a practical tutorial with adaptable R code.

BACKGROUND: The rapidly growing burden of non-communicable diseases (NCDs) among people living with HIV in sub-Saharan Africa (SSA) has expanded the number of multidisease models predicting future care needs and health system priorities. Usefulness of these models depends on their ability to replicate real-life data and be readily understood and applied by public health decision-makers; yet existing simulation models of HIV comorbidities are computationally expensive and require large numbers of parameters and long run times, which hinders their utility in resource-constrained settings. METHODS: We present a novel, user-friendly emulator that can efficiently approximate complex simulators of long-term HIV and NCD outcomes in Africa. We describe how to implement the emulator via a tutorial based on publicly available data from Kenya. Emulator parameters relating to incidence and prevalence of HIV, hypertension and depression were derived from our own agent-based simulation model and other published literature. Gaussian processes were used to fit the emulator to simulator estimates, assuming presence of noise for design points. Bayesian posterior predictive checks and leave-one-out cross validation confirmed the emulator's descriptive accuracy. RESULTS: In this example, our emulator resulted in a 13-fold (95% Confidence Interval (CI): 8-22) improvement in computing time compared to that of more complex chronic disease simulation models. One emulator run took 3.00 seconds (95% CI: 1.65-5.28) on a 64-bit operating system laptop with 8.00 gigabytes (GB) of Random Access Memory (RAM), compared to > 11 hours for 1000 simulator runs on a high-performance computing cluster with 1500 GBs of RAM. Pareto k estimates were < 0.70 for all emulations, which demonstrates sufficient predictive accuracy of the emulator. CONCLUSIONS: The emulator presented in this tutorial offers a practical and flexible modelling tool that can help inform health policy-making in countries with a generalized HIV epidemic and growing NCD burden. Future emulator applications could be used to forecast the changing burden of HIV, hypertension and depression over an extended (> 10 year) period, estimate longer-term prevalence of other co-occurring conditions (e.g., postpartum depression among women living with HIV), and project the impact of nationally-prioritized interventions such as national health insurance schemes and differentiated care models.

Acceptability and feasibility of community-based provision of urine pregnancy tests to support linkages to reproductive health services in Western Kenya: a qualitative analysis.

BACKGROUND: The majority of women living in rural Kenya access antenatal care (ANC) late in pregnancy, and approximately 20% have an unmet need for family planning (FP). This study aimed to determine whether training community health volunteers (CHVs) to deliver urine pregnancy testing (UPT), post-test counselling, and referral to care was an acceptable and feasible intervention to support timely initiation of ANC and uptake of FP. METHODS: We applied community-based participatory methods to design and implement the pilot intervention between July 2018 and May 2019. We conducted qualitative content analysis of 12 pre-intervention focus group discussions (FGDs) with women, men, and CHVs, and of 4 post-intervention FGDs with CHVs, each with 7-9 participants per FGD group. Using a pragmatic approach, we conducted inductive line-by-line coding to generate themes and subthemes describing factors that positively or negatively contributed to the intervention's acceptability and feasibility, in terms of participants' views and the intervention aims. RESULTS: We found that CHV-delivered point of care UPT, post-test counselling, and referral to care was an acceptable and feasible intervention to increase uptake of ANC, FP, and other reproductive healthcare services. Factors that contributed to acceptability were: (1) CHV-delivery made UPT more accessible; (2) UPT and counselling supported women and men to build knowledge and make informed choices, although not necessarily for women with unwanted pregnancies interested in abortion; (3) CHVs were generally trusted to provide counselling, and alternative counselling providers were available according to participant preference. A factor that enhanced the feasibility of CHV delivering UPT and counselling was CHV's access to appropriate supplies (e.g. carrying bags). However, factors that detracted from the feasibility of women actually accessing referral services after UPT and counselling included (1) downstream barriers like cost of travel, and (2) some male community members' negative attitudes toward FP. Finally, improved financial, educational, and professional supports for CHVs would be needed to make the intervention acceptable and feasible in the long-term. CONCLUSION: Training CHVs in rural western Kenya to deliver UPT, post-test counselling, and referral to care was acceptable and feasible to men, women, and CHVs in this context, and may promote early initiation of ANC and uptake of FP. Additional qualitative work is needed to explore implementation challenges, including issues related to unwanted pregnancies and abortion, the financial burden of volunteerism on CHVs, and educational and professional supports for CHVs.

Factors influencing choice of skilled birth attendance at ANC: evidence from the Kenya demographic health survey.

BACKGROUND: In Kenya, skilled attendance at delivery is well below the international target of 90% and the maternal mortality ratio is high at 362 (CI 254-471) per 100,000 live births despite various interventions. The preventative role of skilled attendance at delivery makes it a benchmark indicator for safe motherhood. METHODS: Maternal health data from the Service Provision Assessment Survey, a subset of the 2010 Kenya Demographic Health Survey was analyzed. Logistic regression models were employed using likelihood ratio test to explore association between choice of skilled attendance and predictor variables. RESULTS: Overall, 94.8% of women are likely to seek skilled attendance at delivery. Cost, education level, number of antenatal visits and sex of provider were strongly associated with client's intention to deliver with a skilled birth attendant at delivery. Women who reported having enough money set aside for delivery were 4.34 (p < 0.002, 95% CI: 1.73; 10.87) times more likely to seek skilled attendance. Those with primary education and above were 6.6 times more likely to seek skilled attendance than those with no formal education (p < 0.001, 95% CI: 3.66; 11.95). Women with four or more antenatal visits were 5.95 (p < 0.018, 95% CI: 1.35; 26.18) times more likely to seek skilled attendance. Compared to men, female providers impacted more on the client's plan (OR = 2.02 (p < 0.014, 95% CI: 1.35; 3.53). CONCLUSION: Interventions aimed at improving skilled attendance at delivery should include promotion of formal education of women and financial preparation for delivery. Whenever circumstances permit, women should be allowed to choose gender of preferred professional attendant at delivery.

A case-control study of risk factors for fetal and early neonatal deaths in a tertiary hospital in Kenya.

BACKGROUND: It is important to understand the risk factors for fetal and neonatal mortality which is a major contributor to high under five deaths globally. Fetal and neonatal mortality is a sensitive indicator of maternal health in society. This study aimed to examine the risk factors for fetal and early neonatal mortality at the Moi Teaching and Referral Hospital in Kenya. METHODS: This was a case-control study. Cases were fetal and early neonatal deaths (n = 200). The controls were infants born alive immediately preceding and following the cases (n = 400). Bivariate comparisons and multiple logistic regression analyses were undertaken. RESULTS: The odds of having 0-1 antenatal visits relative to 2-3 visits were higher for cases than controls (Adjusted Odds Ratio (AOR) = 4.5; 95% CI: 1.2-16.7; p = 0.03)). There were lower odds among cases of having a doctor rather than a midwife as a birth attendant (AOR = 0.2; 95% CI: 0.1-0.6; p < 0.01). The odds of mothers having Premature Rupture of Membranes (AOR = 4.1; 95% CI: 1.4-12.1; p = 0.01), haemorrhage (AOR = 4.8; 95% CI: 1.1-21.9; p = 0.04) and dystocia (AOR = 3.6; 95% CI: 1.2-10.9; p = 0.02) were higher for the cases compared with the controls. The odds of gestational age less than 37 weeks (AOR = 7.0; 95% CI 2.4-20.4) and above 42 weeks (AOR = 16.2; 95% CI 2.8-92.3) compared to 37-42 weeks, were higher for cases relative to controls (p < 0.01). Cases had higher odds of being born with congenital malformations (AOR = 6.3; 95% CI: 1.2-31.6; p = 0.04) and with Apgar scores of below six at five minutes (AOR = 26.4; 95% CI: 6.1-113.8; p < 0.001). CONCLUSION: Interventions that focus on educating mothers on antenatal attendance, screening, monitoring and management of maternal conditions during the antenatal period should be strengthened. Doctor attendance at each birth and for emergency admissions is important to ensure early neonatal survival and avert potential risk factors for mortality.

Risk factors for maternal mortality in a Tertiary Hospital in Kenya: a case control study.

BACKGROUND: Maternal mortality is high in Africa, especially in Kenya where there is evidence of insufficient progress towards Millennium Development Goal (MDG) Five, which is to reduce the global maternal mortality rate by three quarters and provide universal access to reproductive health by 2015. This study aims to identify risk factors associated with maternal mortality in a tertiary level hospital in Kenya. METHODS: A manual review of records for 150 maternal deaths (cases) and 300 controls was undertaken using a standard audit form. The sample included pregnant women aged 15-49 years admitted to the Obstetric and Gynaecological wards at the Moi Teaching and Referral Hospital (MTRH) in Kenya from January 2004 and March 2011. Logistic regression analysis was used to assess risk factors for maternal mortality. RESULTS: Factors significantly associated with maternal mortality included: having no education relative to secondary education (OR 3.3, 95% CI 1.1-10.4, p = 0.0284), history of underlying medical conditions (OR 3.9, 95% CI 1.7-9.2, p = 0.0016), doctor attendance at birth (OR 4.6, 95% CI 2.1-10.1, p = 0.0001), having no antenatal visits (OR 4.1, 95% CI 1.6-10.4, p = 0.0007), being admitted with eclampsia (OR 10.9, 95% CI 3.7-31.9, p < 0.0001), being admitted with comorbidities (OR 9.0, 95% CI 4.2-19.3, p < 0.0001), having an elevated pulse on admission (OR 10.7, 95% CI 2.7-43.4, p = 0.0002), and being referred to MTRH (OR 2.1, 95% CI 1.0-4.3, p = 0.0459). CONCLUSIONS: Antenatal care and maternal education are important risk factors for maternal mortality, even after adjusting for comorbidities and complications. Antenatal visits can provide opportunities for detecting risk factors for eclampsia, and other underlying illnesses but the visits need to be frequent and timely. Education enables access to information and helps empower women and their spouses to make appropriate decisions during pregnancy.

A retrospective analysis of maternal and neonatal mortality at a teaching and referral hospital in Kenya.

OBJECTIVE: To measure the incidence of maternal and early neonatal mortality in women who gave birth at Moi Teaching and Referral Hospital (MTRH) in Kenya and describe clinical and other characteristics and circumstances associated with maternal and neonatal deaths following deliveries at MTRH. METHODS: A retrospective audit of maternal and neonatal records was conducted with detailed analysis of the most recent 150 maternal deaths and 200 neonatal deaths. Maternal mortality ratios and early neonatal mortality rates were calculated for each year from January 2004 to December 2011. RESULTS: Between 2004 and 2011, the overall maternal mortality ratio was 426 per 100,000 live births and the early neonatal mortality rate (<7 days) was 68 per 1000 live births. The Hospital record audit showed that half (51%) of the neonatal mortalities were for young mothers (15-24 years) and 64% of maternal deaths were in women between 25 and 45 years. Most maternal and early neonatal deaths occurred in multiparous women, in referred admissions, when the gestational age was under 37 weeks and in latent stage of labour. Indirect complications accounted for the majority of deaths. Where there were direct obstetric complications associated with the delivery, the leading cause of maternal death was eclampsia and the leading cause of early neonatal death was pre-mature rupture of membranes. Pre-term birth and asphyxia were leading causes of early neonatal deaths. In both sets of records the majority of deliveries were vaginal and performed by midwives. CONCLUSION: This study provides important information about maternal and early neonatal mortality in Kenya's second largest tertiary hospital. A range of socio demographic, clinical and health system factors are identified as possible contributors to Kenya's poor progress towards reducing maternal and early neonatal mortality.