The fundamental aim of the study was to identify the household, community, and health system factors that contributed to the child deaths that occurred from 2007 to 2010 in Doume, Nguelemendouka, and Abong–Mbang districts of Cameroon.

Information on births and deaths came from the complete birth histories recorded for all interviewed mothers in a baseline census of all a 16 954 households in the three districts undertaken by Population Services International (PSI) from October to December 2010 for a Home–Based Management of Malaria project.

The sampling methodology of the verbal/social autopsy (VASA) study has been fully described elsewhere [11]. In brief, the study universe included 930 deaths of young children (1–59 months of age) from 2007 to 2010 identified by the census birth histories conducted in the last quarter of 2010. The sampling strategy was to minimize the recall period by taking the one most recent under–five years old death in each household with at least one such death in the four years prior to the census, moving back in time over the survey period until the desired sample size of deaths of 660 child deaths was achieved.

The description of the data collection tools and the fieldwork is available in a paper published earlier [11]. The VASA questionnaire chronologically blended the Population Health Metrics Research Consortium (PHMRC) verbal autopsy questionnaire to determine the biomedical cause of death, with the CHERG Pathway Analysis SA questionnaire [12] to inquire about well–child and illness events leading up to a death. The VASA questionnaire was developed in English and, for the study in Cameroon was translated to French, which is understood by the majority of persons in the study area. Only the local terms for key questionnaire items, such as illness signs and symptoms and the names of local traditional and formal health care providers, were phonetically transliterated to six major local languages– Mongo–Ewondo, Maka, Baka, Mpoong moon, Onveng and Abakoum. The translations were then inserted into a CSProX software application (Serpro S.A, Santiago, Chile) that was developed to enable direct, field–based Computer Aided Personal Interview (CAPI) capture of the VASA interview data on a netbook computer.

For the fieldwork, twenty female interviewers who were native speakers of the local languages and had at least a high school education, received 10 days of in–classroom training in the VASA study background, procedures, ethical standards and conduct of the interview on the netbook, followed by 3 days of field practice, all conducted in French and the local languages. The interviewers were split into three groups (one per district) based on their knowledge of the districts and local languages, and their prior involvement during the mortality survey conducted by PSI in 2010. Each team was led by one field supervisor from the National Statistics Institute (NIS) and in addition received two field visits by office supervisors during the forty days of data collection. The interviewers were trained to select as the respondent the person most knowledgeable of the child’s fatal illness and care provided to the child for the illness. The interview covered the fatal illness from onset to death, including for neonatal deaths, the mother’s pregnancy and delivery. Hence, additional eligible respondents were permitted if necessary. In cases with discordant responses among respondents, the main respondent’s answers were considered. Data collection occurred from 5 March to 15 April 2012.

The analysis of data on preventive and curative care followed the same procedures as described in a prior paper [13], and was guided by the following: (a) review of several sociodemographic and household determinants of the deceased children; (b) coverage of key interventions along the continuum of normal child care provided both inside and outside the home; and (c) illness recognition and care–seeking patterns encompassed by the Pathway to Survival model [9,12,14]. All the examined interventions have been shown to be efficacious and effective in promoting child survival and thus are included among the interventions examined by the Lives Saved (LiST) tool [12] or recommended by the World Health Organization (WHO), and so should be accessible to all children. The list and definitions of some operational variables used throughout this paper are provided in Koffi et al. [13].

In addition, a scoring system was developed based on caregivers’ reports of the child’s feeding behavior, activity level and mental status in order to assess the impact of perceived illness severity at illness onset on caregivers’ attempts at care–seeking for their child’s illness. Hence, three independent categories of illness severity were constructed: normal/mild, moderate, and severe. Details of the method were provided in a prior paper [11]. The Cronbach’s alpha coefficients [15] of the summated scores showed values of 0.90 at the onset of the fatal illness and when the decision to seek care was made. This suggested that the items in the scores elicited highly consistent responses, justifying the reliability of the summated scores according to Nunnaly criteria [16].

Separate to that scoring system, we derived a symptom severity scoring system based on the caregivers observed symptoms by using the World Health organizations’ (WHO) Integrated Management of Childhood Illnesses (IMCI) severity grading for the first symptoms as observed. For the illness symptoms that were in the VA instrument but not in the IMCI, two physician authors (HDK, AKK) assigned symptoms as severe (requiring referral to higher level formal care) or possibly severe (requiring formal health care). The listing of the symptoms and their severity scoring are given in Online Supplementary Document(Online Supplementary Document).

The VASA study in Cameroon was first approved by the Cameroon National Research Committee, then by the Johns Hopkins School of Public Health’s Institutional Review Board. All respondents provided informed consent before the interview was conducted.

Interviews were completed for 635 (96.2%) of the 660 child (1–59 months of age) deaths included in the study sample. More than two–thirds of the respondents (74.7%) were mothers, while 10.6% were fathers of the deceased children, 8.7% grand–mothers, and 6% others relatives.

The sociodemographic characteristics of the deceased children are presented in Table 1 . The median age at illness onset was 12 months (interquartile range IQR:7–24 months), with two–thirds of the deaths occurring either in the post–neonatal (1–11 months of age) period (41.1%) or second–year (26.3%) of life. Median illness duration was 7 days. There were slightly more deaths of females than males, with a male ratio of 96.0. Most (68.8%) of the 635 deceased children were born at home; the majority (58.0%) also died at home.

Table 2 shows the characteristics of the mother, her domestic partner, and the household. Approximately 80% of the mothers were married or living with a man at the time of the interview; two–thirds (67.7%) entered in union before 20 years of age. About a third of the mothers (29.3%) lost their index child before reaching 20 years of age. More than two–thirds (71.2%) had some primary level of education (1–6 years of schooling). The average household size was 7.4 persons. The occupation most cited for the breadwinner was farmer/agricultural worker (68.3%). About a quarter or less of the households had modern facilities such as electricity, an improved source of drinking water, and sanitation (flush or improved pit toilet). The vast majority (96.1%) of the households used firewood as fuel for cooking. The median travel time to the caregiver’s usual health care center was 30.0 minutes. The median time families had been living continuously in the same community was about 10 years.

Figure 1 presents a summary of the nutritional intake before the illness began of the 446 children whose fatal illnesses started between 0–23 months of age. Overall, just 36.3% (n = 162) were being appropriately fed for their age (see Figure 1 ). In more detail, only 15.5% (n = 18) of the 116 children whose fatal illnesses began at 0–5 months of age were exclusively breastfed. Among the 330 children whose illness began at 6–23 months of age, only 32.1% (n = 106) of breastfed children received the recommended complementary non–liquid feeds each day, while 11.5% (n = 38) of non–breastfed children received at least four replacement feeds each day.

Figure 1.  Appropriate feeding for children whose illness started at age 0–23 months. *Child’ illness began before 6 months of age (1-5 months), he/she was being breastfed at the time of fatal illness and was not given anything but breast milk as food. **Breastfed children whose fatal illness started at 6-8 months old and 9-23 months old, and received, respectively, at least two and three complementary non-liquid feedings each day. ***Never-breastfed children whose fatal illness started at 6-8 months old and 9-23 months old, respectively, and received at least four replacement feeds each day (including milk and solid, semi-solid and soft foods). ****Children whose fatal illness started at 0-23 months and satisfied one of the above conditions.

Figure 2 shows some preventive home care received by the 1–59 months old children along the continuum of care. About one in five children (20.8%) were likely not to be exposed to smoke, ie, he/she was usually away from the mother when she cooked inside the house. Less than half (46.5%) of the children always slept under an insecticide–treated bed net before their fatal illness began.

Figure 2.  Coverage along the continuum of care for 1-59-month old child deaths in Doume, Nguelemendouka and Abong-Mbang districts, in Eastern Region of Cameroon, from 2007-2010. *Proportion of children who were NOT usually beside or carried by their mother when she cooked inside the home. **Insecticide-treated bed net. ***Information on immunizations was obtained either from the vaccination card or when there was no written record, from the respondent (mainly the mother). Polio0 is the Polio vaccination given at birth; Fully Immunized children received BCG, measles, and three doses each of DPT and polio vaccine (excluding polio vaccine given at birth).

Figure 2 further shows the percentage of deceased children 12–59 months of age (n = 372) who received vaccinations against each of the six major preventable childhood diseases by one year of age. Overall, just 36.8% of the children were fully immunized against these diseases before they reached their first birthday. The highest coverage was for Polio 0, BCG, DPT1, and Measles, ranging from 86.0% to 94.4%. The deceased children were least likely to be fully immunized against polio by age one (just 52.7% had had all three doses).

Almost all (92.3%) of the 539 6–59–months–old children received at least one dose of vitamin A before the fatal illness began.

The breakdowns in the Pathway to Survival that contributed to the deaths of the children are presented in Figure 3 . This analysis included only the 614 children whose caretakers provided information on the types of actions taken for the illness.

Figure 3.  The “Pathway to Survival” for 614 Young Child deaths in Doume, Nguelemendouka and Abong-Mbang districts, in Eastern Region of Cameroon, from 2007-2010. ¶Median values are reported for the age at illness onset, the delay to formal care, and the illness duration due to the skewed values for these variables. §Illness severity at onset. §§Illness severity at onset and when caregiver decided to seek formal care. N/M=normal/mild, Mod=moderate, Svr=severe *CHWs - Trained Community Health worker. **DK Don’t know.

Nearly all (96.1%) of the caregivers of the 614 children recognized that their child had a severe or possibly severe sign or symptom when they first noticed that the child was ill. Care was provided or sought for almost all (95.8%, n = 588) of the children; while 15 (2.4%) children were said to have “died immediately,” and no care was given or sought for the other 11 (1.8%).

The first action taken for about half (51.9%, n = 305) of the 588 children for whom care was given or sought was to seek care outside the home; the other 283 children (48.1%) first received care inside the home, and 242 of these 283 later sought or tried to seek care outside the home. In total then, 547 (89.1%) of the 614 children for whom care–seeking data was available received, sought, or tried to seek care outside the home. When care was sought outside the home, the majority (74.6%, n = 408) received or tried to seek only formal care, ie, care provide by or at either one of the followings: a trained community health worker (CHW), private doctor or NGO/Government center/post or hospital, 106 (19.4%) received or tried to seek both informal, such as care from a traditional healer or from a pharmacist/drug seller, and formal care, and 33 (6.0%) received informal care only. The median duration (delay) from the illness onset until seeking formal care was 2 days (IQR: 1–3 days) both for those who sought or tried to seek both informal and formal care and those who sought or tried to seek only formal care. For both groups as well, the decision to seek formal care was delayed by 2 days (median time) after the onset of the illness ( Figure 4 ), regardless of whether formal care was sought from a health worker in the community or at a hospital or other health facility (χ2(2) = 1.261; P = 0.2614). In addition, the median delay was 2 days among children who were perceived to be normal or moderately ill, compared to the median of 2.5 days among those who were severely ill when the decision was made to seek formal care. A median test showed that there was no statistically significant difference in delay between the normal/moderately ill and the severely ill children (χ²(1) = 0.831, P = 0.3619).

Figure 4.  Illness severity ranking at onset and at decision to seek care among children for whom caregivers tried to seek or sought some formal care (N = 506). 8 children had missing information that did not allow their illness severity ranking.

Among those who sought or tried to seek some formal care (N = 514, including 8 who had missing information on perceived illness severity), the percent of children perceived to be severely ill increased from 24.1% (122 out of 506) at the time of illness onset to 43.5% (or 220 out of 506) when the decision was made to seek formal care ( Figure 4 ). Many (n = 111) children who were mildly or moderately ill at the time of illness onset, became severely ill by the time their caregivers decided to seek formal care. In other words, the mean illness severity score increased from 1.66 (±SD = 1.043) at onset of the illness to 2.15 (±SD = 0.923) when the decision was made to seek formal care, and the difference of –0.49 (±SD = 0.893) was statistically significant (t = –12.35; P < 0.0001).

Of the 514 children for whom caregivers tried to seek some formal care, 45 (8.8%) did not reach the health care facility because they died before setting out, died en route or could not reach the health care provider. The remaining 469 (91.2%) children reached the first health care provider after about 30 minutes median travel time, IQR:15–60–minute. Thirty–one (31) went to a community health worker (CHW), 10 to a health post, 22 to a private doctor or clinic, 178 to an NGO or government clinic, and 226 to an NGO or government hospital, and 2 – for which the name or type could not be identified with the available data.

Out of 469 children that reached a first provider, 101 (21.5%) died at that provider. Approximately half (51.5%, or n = 52) of those children who died at the first provider were judged by their caregivers to be severely ill at the time the decision was made to seek formal care. This compares to slightly more than a third (38.6% or 142 out of 367) of those children who left the first provider alive being judged to be severely ill at the time the decision was made to seek formal care. The difference between the two groups (51.5% vs 38.6%) was statistically significant (χ²₁ = 9.325; P < 0.010.

In addition, about 39% (n = 144) of the 368 that reached a health care provider and left the provider alive were not referred nor given any home care recommendations. The remaining 224 were either only referred (n = 20) to a second health care provider, only received home care recommendations (n = 157), or were referred and received home care recommendations (n = 47). In summary, just 67 (18.2%) of the 368 that left the first provider alive were referred; however, when recommendations were received, or referrals provided, most of the caregivers (77%–82%) followed all the recommendations or accepted the referral and went to a second health care provider.

Figure 5 explores the care–seeking constraints for fatal child illnesses. In total, 400 of the 588 caregivers (68.0%) whose children received, sought or tried to seek care reported that they had some concerns or problems in seeking care from a health care provider for their child’s fatal illness. Cost (82.3%), lack of transport (24.3%) and distance (22.8%) were the primary constraints for care–seeking at a health provider, with more caregivers who did not seek care than those who did seek care reporting that they had a concern or problem.

Figure 5.  Main care–seeking constraints for child illness (N = 400 caregivers).

In conclusion, as the global health community deliberates the strong likelihood that MDG–4 targets will not be accomplished and about the post–MDG era, it is important that aspects that have been ignored over the past decade find emphasis and support both nationally and globally. The recent social autopsy study conducted in Doume, Nguelemendouka, and Abong–Mbang health districts in the Eastern Region of Cameroon sheds light on the most common household, community and health care system factors that contributed to the deaths of children under five years of age. Among these factors are poor living conditions, poor nutritional status, prevailing customs or cultural practices that lead to exposure to indoor smoke, and health–related behaviors such as delaying the decision to seek care.

Short–term interventions could include the introduction of the C–IMCI program that could increase caregivers’ ability to recognize danger signs of child illnesses and facilitate behavior change for timely and appropriate health care–seeking. Building informed demand among children’s caregivers to seek prompt treatment from an appropriate provider is an important component of any intervention aimed at improving case management coverage. In addition, keeping improved infant and young child feeding high on the public health agenda is crucial to consolidating gains made during the past two decades.

An improved standard of living such that parents or caretakers can overcome the economic obstacles to seeking basic child health care might be efficacious in the long term.