Study design

Deterministic mathematical modeling was done using Lives Saved Tool (LiST) and Missed Opportunity Tool (MOT) [12]. A demographic module (DemPro), family planning module (FamPlan), and AIDS impact module (AIM) were used as part of the LiST analysis. These tools are part of a suite of policy modeling tools in the spectrum software available at www.avenirhealth.org. Version 6.08 was used.

The LiST is a multi-cause deterministic mathematical population-level modeling tool used for evaluation and attribution. It can be used to evaluate the impact of multiple interventions or programmes implemented simultaneously. There were over 70 interventions in LiST at the global level as of 2011. The LiST has been validated and used in studies in several countries including Ghana [13–19]. It was used to model the number of lives saved and percent mortality rate reduction based on changes in intervention coverage levels and the decrease in the prevalence of stunting and wasting in this study [20–23].

In general, if coverage levels increase for the interventions, children who were previously unprotected from death due to lack of use of these interventions become protected. Therefore, additional lives are saved, or excess deaths are prevented due to the interventions relative to the previous coverage levels. Conversely, if coverage levels are reduced, it results in excess deaths, while no change in coverage results in zero lives saved. The interventions used in this analysis were those with coverage level information in the LiST. For stunting and wasting, a decrease in their prevalence reduces the risk of infections such as malaria, diarrhoea, and pneumonia and, consequently, deaths from these conditions.

The MOT determines the interventions that will make the most impact if intervention coverage levels were individually scaled up to 90%, one at a time, one year from the baseline year [12]. It, therefore, estimates the future potential impact of interventions, and thus, those to focus on to accelerate mortality decline. Although 90% is the default coverage level in the missed opportunity tool, it is also the target coverage level of vaccines and other interventions for the Ghana Health Service [24].

Sources of data

The lives saved tool contains default data of several countries from different sources, including household surveys such as the Demographic and Health Surveys (DHS), Multiple Indicator Cluster Survey (MICS), and Malaria Indicator Survey (MIS), the World Health Organization (WHO), and the United Nations Children Emergency Fund (UNICEF). Coverage of maternal and child health intervention data comes from surveys such as DHS, MICS, and MIS, apart from vaccines from the WHO/UNICEF. Water and sanitation-related interventions come from the WHO/UNICEF Joint Monitoring Program for water supply and sanitation (WHO/UNICEF-JMP). Mortality rate data comes from the UN Inter-agency Group for Child Mortality Estimation (IGME). Coverage levels of some interventions for which data is not available are modeled using other intervention coverage levels, or assumptions are made on their coverage levels. Appendix A contains interventions in LiST on which data is available in Ghana.

Lives saved tool modeling strategy

The lives saved tool makes a projection of the population under study using the DemPro (demographic projection) and calculates births and deaths that should occur in that population. The effect of family planning services on the number of births is applied using information from the FamPlan module. At the same time, the effect of AIDS on mortality is also applied using information from the AIM (AIDS impact module). With the number of deaths estimated, coverage of interventions to reduce mortality is applied to the population. Based on the effectiveness of the intervention(s) delivered in that population, the projected deaths are expected to change relative to the changes in coverage of the interventions since the interventions are expected to reduce the number of deaths that should occur in the population.

The basic formula for calculating the impact is impact = change in intervention coverage X effectiveness of intervention X affected fraction. Adjustments are made to this basic formula to account for previous lives saved and the herd effect of vaccines and bed net. Details of the modeling strategy are built in the LiST and published [25].

Intervention effectiveness is effectiveness established at the global or regional levels. It is in the LiST, while the affected fraction is the fraction of cause-specific mortality amenable to the particular intervention which is also in the tool [25].

The timeframe for the modeling was from the year 2008 to 2014. 2008 and 2014 were chosen because they are closest to the end of the Millennium Development Goals (MDGs) and the beginning of the Sustainable Development Goals (SDGs). Therefore, information from the study will provide an understanding of which interventions to focus on during the SDGs period to achieve rapid mortality reduction among children under five years old. The definitions of the interventions evaluated in this study are found in the LiST.

Missed opportunity tool modeling strategy

The MOT estimates lives saved by automatically scaling up interventions with coverage levels below 90% to 90%, one intervention at a time. Interventions with coverage levels at or greater than 90% are excluded. Interventions with the most lives saved are those likely to contribute to lives saved/mortality reduction if their coverage levels are increased.

Study area

The study was conducted in Ghana, a country with stable democratic governance since 1992. Healthcare expenditure in 2014 was 60 United States Dollars per capita [26]. Schieber, Cashin [10] has documented that Ghana’s U5M rate is high compared to its high healthcare spending. Related to this is Ghana’s slow progress made on under-five mortality during the period of the Millennium Development Goals (MDGs) [6].

Study variables

Information in the LiST used were mortality rates (neonatal, infant, and under-five), coverage of interventions, stunting, wasting, and contraceptive prevalence. The outputs were mortality rates (neonatal and under-five) and lives saved. Neonatal mortality rate is the probability of dying within the first month of life. In contrast, under-five mortality is the probability of dying between birth and exactly five years [3]. The interventions included were iron intake, intermittent preventive treatment of malaria in pregnancy, tetanus toxoid vaccine (neonatal tetanus protection), skilled delivery, clean postnatal care, early initiation of breastfeeding, water connection in the home, time to improved water source, improved sanitation and insecticide-treated net/indoor residual spraying. From the LiST, at baseline (2008), the neonatal mortality rate was 31.2 deaths per 1,000 live births, while the under-5 mortality rate was 75.3 deaths per 1,000 live births. At the end line (2014), the neonatal mortality rate was 26.6 deaths per 1,000 live births, while the under-5 mortality rate was 57.1 deaths per 1,000 live births in Ghana.

Definition of some of the interventions

Caesarean delivery: percent of women requiring surgical intervention if indicated (emergency caesarean birth for obstructed labour, uterine rupture, or foetal distress).

Complementary feeding: percent of 6–23-month-old children receiving minimum dietary diversity (4+ food groups).

Case management of severe neonatal infection (sepsis/pneumonia): sum of the three levels of case management for severe infection in the neonatal period: oral antibiotics, injectable antibiotics, and full supportive care.

Full supportive care for neonatal sepsis/pneumonia: full supportive care for neonatal sepsis/pneumonia.

Artemisinin compounds for the treatment of malaria: percent of children treated within 48 hours of the onset of fever in malaria-endemic areas with an artmesinin-containing compound (artemisinin-based combination therapy).

Data analysis

The default data of Ghana in the LiST was loaded into the tool, and the analyses done. First of all, to assess the contribution of only interventions to mortality decline between 2008 and 2014, the baseline for the impact evaluation was set at 2008 and the end line in 2014. Then, interpolation was done between 2008 and 2014 (linear scale-up of coverage of interventions between the baseline and end-line), and lives saved by interventions between 2008 and 2014 were reported. Finally, changes were made to this configuration (modeling strategy) to allow for estimates of lives saved, and mortality rate reduced based on the direct entry of stunting and wasting (reduction in the prevalence of stunting and wasting between 2008 and 2014) in addition to the interventions. Lives saved due to changes in intervention coverage and differences in the prevalence of stunting and wasting were reported.

Secondly, to assess the interventions with the most potential to reduce mortality if coverage levels were scaled up individually to 90% from the 2014 coverage levels, a new projection was created with 2014 as the baseline and 2019 (a minimum of 5 years is allowed to estimate impact using the LiST) the end line. The 2014 baseline projection was then uploaded into the missed opportunity tool, and lives saved by each intervention were reported. All analyses were done in duplicates, and the results were the same.

Lastly, for the analysis to assess the impact of interventions at scale-up of coverage levels to 100% in 2014, the projection created with 2008 as the baseline and 2014 as the end line was used. Next, all the coverage levels of interventions with available data in 2014 were changed to 100%, interpolation done and lives saved, mortality reduction and neonatal and under-five mortality rates reported. One hundred percent (100%) was chosen because coverage levels of DPT, pneumococcal, rotavirus, and measles vaccines in the LiST were over 90%.

Results

A total of 48,084 lives were saved (Table 1), resulting from changes in coverage levels of interventions, and reduced stunting and wasting among children born between 2003 and 2014. In addition, modeling involving only the interventions resulted in 34,477 lives saved/additional deaths prevented.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Additional child lives saved and mortality reduction by intervention between 2008 and 2014 among children 0–59 months.

https://doi.org/10.1371/journal.pone.0267776.t001

Interventions that saved the most lives considering the effect of reduction in the prevalence of stunting and wasting among children under-five years were ITN/IRS, 6,437 (13.4%) lives saved, ACTs, 4,325 (9.0%) lives saved, case management of neonatal sepsis/pneumonia, 4,271(8.9%) lives saved, caesarean delivery, 4,145(8.6%) lives saved, and pneumococcal vaccine, 3,938 (8.2%) lives saved (Table 1). Reduction in wasting saved 10,372 (21.6%), while a decrease in the prevalence of stunting also saved 4,315 (9.0%) lives. On the contrary, breastfeeding, -1,786(-3.7.0%), piped water, -190 (-0.4%), antibiotics for treatment of dysentery, -118 (-0.2%), and vitamin A supplementation, -21 (-0.0%) resulted in negative lives saved (excess deaths/additional deaths) between 2008 and 2014 (Table 1).

If coverage levels of interventions in 2014 were scaled-up to 90% in 2015, each intervention at a time, full supportive care for prematurity (5,435 lives saved), full supportive care for neonatal sepsis/pneumonia (3,002 lives saved), assisted vaginal delivery (2,163 lives saved) and kangaroo mother care (2,010 lives saved) would have saved the most lives among neonates (Table 2). In addition, ACTs (4,925 lives saved), ORS (2,056 lives saved), antibiotics for the treatment of pneumonia (1,805 lives saved), and prevention of mother-to-child transmission of HIV (including breastfeeding choices) (1,020 lives saved) would have made the most impact on mortality reduction among children 1–59 months (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Lives saved at 90% intervention coverage level, one intervention at a time among children 0–59 months.

https://doi.org/10.1371/journal.pone.0267776.t002

Also, if coverage levels of all the interventions were scaled up to 100% in 2014, the neonatal mortality rate would have been 21.7 deaths per 1,000 live births instead of 26.2 deaths per 1,000 live births, while the U5M rate would have been 40.1 deaths per 1,000 live births instead of 57.1 deaths per 1,000 live births in 2014. Under this scenario (coverage of interventions at 100%), results from the effect of only interventions showed that ACTs, 21,204 (19.6%), ITN/IRS, 12,906 (11.9%), oral antibiotics for pneumonia, 9,173 (8.5%) and age-appropriate breastfeeding practices, 7,935 (7.3%) would have contributed the most to mortality reduction (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Additional child lives saved by intervention in 2014 intervention coverage levels were scale-up to 100% among children 0–59 months.

https://doi.org/10.1371/journal.pone.0267776.t003

Discussion

Ghana failed to achieve the MDG4 target despite efforts at increasing intervention coverage levels. This study, therefore, evaluated the contribution of the various interventions to mortality decline. Information from the study will help prioritize intervention scale-up. Reduction in the prevalence of wasting and stunting, malaria control interventions (ACT and ITN/IRS), pneumococcal vaccines, and childbirth-related interventions (case management of neonatal sepsis/pneumonia, caesarean delivery, clean cord care, and thermal protection) caused mortality reduction. In contrast, nutrition (breastfeeding, vitamin A supplementation, and vitamin A for the treatment of measles) and water-related (piped water and point-of-use filtered water) interventions were responsible for excess mortality between 2008 and 2014.

Proper maternal and child nutrition plays a crucial role in child survival. It reduces the risk of important causes of death such as malaria, diarrhoea, and pneumonia [6,7] and risk factors such as preterm, small for gestational age, stunting, and wasting [3,27]. Unfortunately, breastfeeding and complementary feeding prevalence reduced between 2008 and 2014 among children under-five years in Ghana [3,28]. This resulted in -1,786 (-3.7%) excess deaths due to a reduction in breastfeeding (60.6% to 50%) between 2008 and 2014. Despite this, a decline in stunting (28.0% to 18.8%) and wasting (8.5% to 4.7%) was achieved between 2008 and 2014, contributing to a 9.0% and 21.6% mortality reduction, respectively. The food fortification programme with vitamin A, the Livelihood Empowerment Against Poverty (LEAP) programme introduced in 2007, the school feeding programme introduced in the 2005–2006 academic year [29], and poverty reduction during the period of evaluation could explain the decrease in stunting and wasting despite a reduction in the prevalence of complementary feeding [3,28]. It implies that further mortality decline can be achieved if complementary feeding and breastfeeding coverage levels are increased. Similar results of reducing stunting and wasting to mortality decline have been reported [13,30–32].

Generally, negative lives are saved if the coverage level of an intervention is reduced during the evaluation period. This means taking the population that needed to be protected by the intervention, the percentage protected was less at the end line than the baseline. Therefore, a proportion of the people became unprotected and died as a result. It could be the case that the intervention is available but not all the population that needs it uses it. Examples of such interventions are breastfeeding, vitamin A supplementation and antibiotics to treat dysentery. The negative lives saved could also be due to an increase in the population size not matched with the scale-up of the intervention. This could be the case with piped water. Population growth is not commensurate with the provision of piped water.

Malaria remains a top cause of under-five deaths in Ghana, and therefore, effective interventions targeting malaria deaths should have a high impact. While household ITN/IRS coverage made a substantial impact (15% mortality reduction), they still have the potential for increased impact if coverage is scaled up. Similar results of the effect of bed net on mortality have been documented [13,17,30,33]. Comparing ACTs to ITN/IRS, ACTs also saved many lives (9% reduction in mortality rate), but this was lower than that achieved by ITN/IRS (13.4% reduction in mortality rate). This is likely due to the low coverage change of ACTs (11.3% to 26.6%) between 2008 and 2014 compared to ITN/IRS (41.7% to 70.9%). This low coverage could explain the continuously enormous contribution of malaria to U5M and the generally low mortality decline in Ghana since uncomplicated malaria can progress rapidly to severe disease and death due to delay in treatment. It also implies that ACT can save additional lives since its current coverage level is low in the country. Therefore, it should be one of the interventions to focus on to further reduce mortality, especially in malaria high burdened regions of Ghana.

Vaccines have historically, and in recent times, contributed greatly to mortality reduction [17,30,31,34]. In this study, pneumococcal and rotavirus vaccines greatly impacted mortality reduction, 8% and 3%, respectively. These vaccines are relatively new, were introduced in 2012, accounting for their high coverage level change, 0 to 99% for pneumococcal vaccine and 0 to 98% for rotavirus vaccine between 2008 and 2014 [3,8]. However, they are not likely to save more lives than their current impact in the future, just like the old vaccines. This is because, with a national level coverage of over 90% for most vaccines, any further increase in coverage will be less than 10%. Also, if the high coverage levels of these vaccines reduce the prevalence of vaccine-preventable diseases in the population like diphtheria and tetanus, they will likely not contribute greatly to mortality decline even if their coverage levels are further increased since very few children will suffer from these conditions. Considering that one of the most effective tools for reducing mortality does not have much potential for further mortality reduction, coverages of other high-impact interventions, especially those with low coverage levels need to be increased. The recently introduced malaria vaccine might contribute to filling this gap [9].

In this study, interventions related to childbirth, such as caesarean delivery delivered by skilled personnel, contributed the most to lives saved (8.6% of lives saved). Skilled delivery is one of the main strategies to reduce neonatal mortality, an important contributor to U5M. In Tanzania, skilled delivery and emergency obstetric care among neonates contributed to a 29% (6,200 lives saved) reduction in mortality between 2000 and 2012 [30]. However, Ethiopia achieved the MDG4 target on child survival with skilled delivery coverage between 2000 and 2011 of less than 5%, contributing to a 3% decline in U5M [31]. The quality of skilled delivery has been questioned, and neonatal mortality ranked second in deaths due to poor quality of care [35].

In Ghana, while the coverage level of skilled delivery was 84.3% in 2008 and 92.1% in 2014 in the Greater Accra Region, neonatal mortality increased from 21 deaths per 1,000 live births to 25 deaths per 1,000 live births between 2008 and 2014. Also, in the Ashanti Region, skilled delivery coverage was 72.6% in 2008 and 86.3% in 2014; still, the region recorded an increase in neonatal mortality from 35 deaths per 1,000 live births to 42 deaths per 1,000 live birth from 2008 to 2014 [3,28]. On the contrary, in the Upper West Region, where the skilled delivery coverage level was 46.1% in 2008 and 63.7% in 2014, neonatal mortality decreased from 45 deaths per 1,000 live births to 24 deaths per 1,000 live births between the periods. Likewise, in the Northern Region where skilled delivery coverage was 27.2% in 2008 and 36.4% in 2014, neonatal mortality declined from 35 deaths per 1,000 live births to 24 deaths per 1,000 live births within the period [3,28].