Assessment of infant mortality surveillance : case study

Objectives: assess the implementation of child mortality surveillance in Recife/PE. Methods: an analytical evaluative study was conducted on its implementation. It was a single-case study that correlated degree of implementation with the of the result indicators surveillance. A logic model on this strategy and a matrix of indicators and judgments according to model components were drawn up. The degree of implementation was obtained from structure and process indicators and this was then correlated with result indicators, in a deductive approach based on intervention theory. Results: the structure approach presented superior results to the process in all evaluated components. This strategy was considered to have been partially implemented (75.7%), however, the components of ‘identification of deaths’ (85.7%), ‘epidemiological research' (88.1%) and 'referral of proposals for promotion and health care and correction of official statistics' (95.8%) were classified implemented. Regarding the relation of the degree of implantation of the surveillance and its results with the logical model, only one of the 17 indicators was considered inconsistent. Conclusions: this strategy was considered to have been partially implemented. The model of child mortality surveillance and its assessment were shown to be adequate for signaling the consistency of the interrelations between the activities proposed and the effects expected, and would be reproducible within other scenarios.


Introduction
Infant death is a sentinel event of the quality of healthcare given the early and often preventable nature of these events. 1 In view of the magnitude and transcendence of infant mortality, in 2000, the United Nations established the reduction of these deaths by 2/3 by 2015 as one of the Millennium Development Goals. 2 Although Brazil achieved this goal in 2012, some countries did not reach the goal despite their efforts, leading to an extension of the deadline to 2030. 3,4everal nations have acknowledged the need to associate the epidemiological profile of mortality with quality assessments of care provided to women and children during their trajectory, and have perfected strategies to facilitate greater understanding of social and care-related determinants of the deaths, including the mortality surveillance. 5,6n the inspection and investigation of child deaths, international experiences have produced several configurations based on availability of information and the supply of more or less complex health services. 5,7,8The investigation of deaths can explain the circumstances in which they occurred and the results should be notified to managers and any interested people so they may act on various levels of the health system to improve access and quality of care provided during pregnancy, labour, childbirth, and child and mother follow-up. 5,7,8n Brazil, a similar strategy was implemented by the Ministry of Health (MH) in 2010, with the publication of a normative base of Infant Death and Foetal Surveillance, 9 although municipal and state initiatives with different conformations have existed for some years.[10][11][12][13][14][15] In Recife, since 2003, infant mortality surveillance was implemented to investigate and discuss all infant deaths (except those caused by congenital malformations) of mothers living in Brazil.Cases are discussed with healthcare workers, inspectors, and managers in order to identify preventable deaths caused by medical errors, to use this information for reflection and planning, and to adopt measures to reduce infant mortality.10 In the surveillance assessments conducted in Brazil, the strategy is considered a management tool capable of providing information to health teams, generating critical awareness on the offered care, and enhancing information systems.11,16,17 Moreover, it can be used to plan interventions for the main problems and barriers of the care system and to reduce preventable deaths and iniquities.6 Given its potential, mortality surveillance generates special interest for the monitoring actions required to achieve the Sustainable Development Goals.2,3 Despite the relevance and scope of national infant mortality surveillance and the guidelines adopted by Brazilian states and municipalities, no publications assessing the implementation of this strategy in Brazil were identified.A case study, with its high potential for internal validity, provides valuable insight for managers, technicians, and researchers.The aim of this paper is to assess the implementation of infant mortality surveillance in a capital city from northeastern Brazil.

Methods
This is an evaluative analysis of qualitative research of infant mortality surveillance.The methodological strategy is a single case study to relate the degree of implementation to the result indicators of intervention in the city of Recife, 18,19 northeastern Brazil, thestate capital of Pernambuco with the fourth largest urban concentration in the country.The city covers an area of 218.5 km 2 and its population, according to the Brazilian Institute of Geography and Statistics, is estimated at 1,617,260 inhabitants (2015). 20The mother and child healthcare network consists of 17 maternity hospitals, 275 family health teams (60% coverage), 21 family health support centres, 22 health centres, and 12 polyclinics.
A logical intervention model was initially prepared to clarify how the components work together using activities and available resources to achieve the expected results.The following documents were used for the model: infant mortality surveillance implementation project; 21 MS Ordinance No. 116/2009 that regulates data collection, flow and periodicity of data on deaths and live births submitted to the information systems; MS Ordinance N o .72/2010 establishing the obligation of infant and foetal mortality surveillance; and the manual of foetal and infant mortality surveillance and infant and foetal death prevention. 22The listed intervention components were: 'identification of deaths', 'epidemiological investigation', 'discussion of deaths', and 'submission of health promotion and care proposals and correction of official statistics' (Figure 1).
A matrix of indicators and a judgment was built according to the component of the logical model related to the approaches (structure, process, and result).A parameter was established for each indicator based on the legal instruments mentioned above.The non-normalised indicators were defined according to the service routine of the municipality

Components
Inputs Activities Short/Medium Term Results

Table 1 concluded
Matrix of indicators and parameters by component and approaches of infant mortality surveillance in Recife.

Components
Person inputting data at central level is competent to apply with corrections proposed by the Technical -Management Team in the Sinasc and MIS databases % of summary sheets sent from sanitary district to central level Data of 2014 were collected from the MIS, including the online version; Information System on Live Births (Sinasc); municipal system of death certificate (DC) distribution and control; infant mortality surveillance panel; infant death and birth certificate (BC) submission protocol for sanitary districts (SD); form of investigation and summary of infant deaths, death certificate submission sheets at the health units and necropsy services, compulsory notification of death (CND) by disease, and death entry sheets.
One hundred indicators were analysed, of which 83 were used to define the degree of implementation (structure and process) and 17 were used to define the results.The maximum estimated score was 100 points divided into the four components: 'identification of deaths' (10 points), epidemiological investigation' (30 point), 'discussion of deaths' (40 points), and 'submission of health promotion and care proposals and correction of official statistics' (20  points).
A lower score division was assigned to the component 'identification of deaths' because it is part of the MIS, essential for mortality surveillance.For 'submission of health promotion and care proposals and correction of official statistics' a lower score was also attributed because it is a major interface for the discussion on deaths between the workers and managers.The highest scores were assigned to 'epidemiological investigation' and 'discussion of deaths' since they are central aspects of intervention.A weight of 30% was attributed to the approach 'structure' and a weight of 70% was attributed to 'process' because the assessment was performed in a large municipality with a minimally structured services network.
The classification of the degree of implementation of infant mortality surveillance, arbitrated by the authors, was based on the time and duration of intervention, according to the indicators of structure and process, and categorised as: implemented (between 80.0 and 100.0%), partially implemented (60.0 to 79.9%), and not implemented (<60.0%).
Subsequently, the degree of implementation by components and as a whole was confronted with the theory of intervention, expressed in the logical model with empirical results based on deduction, and the intensity of relations was noted as consistent (degree of implementation and result indicators in the same category), partially consistent (degree of implementation and result indicators included in the category immediately above or below), and inconsistent (involves the relationship between the highest and lowest degree of implementation categories and result indicators). 18,23The degree of implementation analysis was confronted with each of the result indicators of respective components.
This study was approved by the research ethics committee of the Centro de Pesquisas Aggeu Magalhães da Fundação Oswaldo Cruz (CAAE: 07336313.6.0000.5190).The interviews were scheduled and recorded after the participants signed an informed consent statement.

Results
In the universe of structure indicators, all the infant mortality surveillance components were above 85%.In the component 'submission of health promotion and care proposals and correction of official statistics', all the indicators obtained a high score, while 'discussion of deaths' reached 85.7% of the expected score (Table 2).
In relation to the process approach shown in Table 3, 'submission of health promotion and care proposals and correction of official statistics' was the best rated (91.7%) and 'discussion of deaths' obtained the lowest score (67.9%).Of these indicators, five were rated zero: percentage of discussions with the participation of representatives of outpatient units and hospitals involved in care, managers/technicians of maternal health and primary care at central level, and sanitary district manager.The structure approach obtained higher results than theprocess in all the assessed components.
Infant mortality surveillance was classified as partially implemented (75.7%).Three components, however, were classified as implemented, namely 'identification of deaths', 'epidemiological investigation', and 'submission of health promotion and care proposals and correction of official statistics', with 85.7%, 88.1%, and 95.8%, respectively.
Of the 17 result indicators in relationship assessment of the degree of implementation of infant mortality surveillance and its results with the logical model, only one indicator was considered inconsistent, namely 'percentage of infant deaths discussed with discussion reportprepared and submitted' (13.6%).This indicator is linked to the component 'submission of health promotion and care proposals and correction of official statistics' (Figure 2).

Discussion
Infant mortality surveillance was considered partially implemented, although three components ('identification of deaths', 'epidemiological investigation', and 'submission of health promotion and care proposals and correction of official statistics') were classified as implemented.The component 'discussion of deaths', which was attributed the highest weight, was decisive in the definition of the degree of implementation obtained in the intervention.Moreover, the empirical results obtained in mortality surveillance were compatible with its degree of implementation.
The structure approach was not considered a problem for operation and intervention activities, although other studies state the scarcity of resources as an obstacle to more productive action. 11,24Some difficulties were detected in the indicators of 'identification' and 'discussion of deaths', all of which are related to failure to notify deaths in a timely manner and low participation of important actors in the discussion.
The identification of a child's death triggers the case notification process.This notification is submitted after the death certificate issued by the health workers and services is sent to municipal departments of health, within 48 hours from the date of occurrence as recommendation. 9This component depends on the quality of the MIS, which has been perfected over the years, and precedes the investiga-    Relationship ofthedegree of implementation of infant mortality surveillance components and their results with the logical model Recife, 2014.tion of infant deaths. 25he rapid flow of information and the timely start of investigations are important for the success of mortality surveillance. 22In Brazil, the deadline to complete an investigation is 120 days from the date of the occurrence, 9 while in the United Kingdom the investigation must be concluded within one month of the death notification. 26Delays in this process can hinder interventions to improve the quality of vital statistics information systems and prevent new deaths.

Identification of deaths
This study identified that almost all (95.1%) infant deaths, except those caused by congenital malformation, were investigated.This percentage is similar to the results found in the city of Londrina -Paraná (97% to 100% of deaths) 15 and different from the results of Arapiraca -Alagoas (54.3%). 17The overall proportion of deaths investigated in Brazilian states is lower (in Bahia 21.3% and in São Paulo 57% of health management units investigated all infant deaths). 11,14Moreover, the manner in which the forms are completed must be improved, especially in terms of investigating the home, the prenatal care and child care units, and the necropsy services.It is critical to improving the quality of investigations in order to build a chain of events that supports death notifications and provides a better understanding of the socioeconomic, cultural, and care determinants. 5,11,17ational and international discussions of deaths occur in committees by a group of experts from universities, health departments, councils, and nongovernmental organizations. 11,27,28The studied municipalitydiffers when it that adds social services, inspection, and management workers into the discussion.Nonetheless, this was the component with the lowest degree of implementation.
The absence of relevant actors from the health system in discussions of deaths has a negative effect on one of the most important activities of this strategy.Discussions enable a change of attitude and social practices, improve the education of healthcare workers and managers, and increase the effectiveness of the recommendations.The participation in discussions and horizontal listening sessions of workers and managers with multiple roles and levels of involvement in the deaths can help establish ways to overcome barriers and provide care that better suits the needs of women and children.A study conducted in a capital city of northeastern Brazil on preventable infant deaths and barriers in primary care revealed conflicting understandings of the event depending on the position of the care networkor whether the person was the child's mother. 29 spite of the strategy's potential, international research on the social autopsy of mother and child deaths conducted in several countries found resistance among health workers to report or discuss these deaths due to fear of being held accountable or penalised. 27nalyses and studies of each death have also helped improve the healthcare information systems since, after the investigations, the team that discussed the case usually attributes new root causes to the infant's death, and completes and/or validates the variables of the live birth and death certificates, 9,12,13,16,17 as observed in this assessment.Furthermore, improving vital information systems promotes changes and adjustments to the child mortality profile and enables appropriate planning of actions for its challenging.
A low proportion of infant deaths involved the preparation and submission of reports with propositions of the healthcare sector to the sectors responsible for taking the necessary precautions.These reports provide an overview of the quality of mother and child care in order to encourage healthcare authorities to act. 27However, the discussion of death with the workers involved in care allows a change of attitude based on their own participation because of the educational, reflective, and purposeful nature of these meetings, which was also observed in perinatal audits in other regions of the world. 28The experience of Moldova shows that if all stages of this strategy are correctly executed, the quality of mother and child care will also improve and possibly reduce the occurrence of preventable infant deaths. 28he consistency between the result indicators and the degree of implementation of infant mortality surveillance suggests the model is suitable to understand the degree of implementation and situation of all activities in the studied case.One of the advantages of making the logical model of an intervention explicit is the possibility of extrapolation to other scenarios.In addition, the model improves the studied strategy by precisely outlining the objective, enhancing individual and collective actions, and identifying planned activities to achieve the expected results in an educational training perspective, thus strengthening the acts and decisions of all stakeholders involved. 19he strengths of the studied strategy are related to the quality and complexity of the logic on which the research is based, and in the plausibility of interrelationships between the activities and the results, which guarantees greater internal validity. 19owever, in terms of external validity, a limit would be the extrapolation of results.The logical model can be extrapolated, not the results. 23n conclusion, infant mortality surveillance does not fully adhere to the norms and guidelines established in regulatory documents.Some factors regarding this point were identified as being more fragile and should be considered in interventions to improve the strategy.In order to overcome the obstacles preventing appropriate mortality surveillance, scientific, organisational, and experience-based knowledge must be correctly combined and applied.Discussing a child's death is a problem situation of circumstance that requires reflection, followed by decoding, acknowledgment, and denaturalisation, especially in situations of restricted budget and rights. 30he logical model of infant mortality surveillance and its assessment were considered appropriate because they were capable of detecting the consistency of interrelationships between the proposed activities, and can, therefore,be reproduced in other scenarios.This type of assessment is critical to identify the weaknesses and potentialities of the strategy and enable managers and technicians to act according to requirements.

Figure 1
Figure 1Logical model of infant mortality surveillance in Recife.

continue
Degree of implementation parameter/score -Results in the proportions 80-100%=1; 60-79%=0.75;40-59%=0.5;20-39%=0.25,and <20%=0.Yes=1 or No=0.-Computer capacity: dual core or above=1; 1 to 2 Gigahertz (GHz) with dual core=0.75; 1 to 2 GHz with single core=0.5;GHz or less with dual core=0.25; 1 GHz or less with single core=0.-Internet capacity: above 15 Megabytes second (Mbps)=1; 11 to 15 Mbps=0.75;6 to 10 Mbps=0.5; 1 to 5 Mbps=0.25; 1 Mbps or less=0.-Meeting record Minute=1 or other document=0 DC=death certificate; BC=live birth certificate; SD=sanitary district; NC=central level; MIS= Mortality Information System; HEU=hospital epidemiology unit; FHS=family health strategy; PACS=Community Health Agent Programme; IMS=infant mortality surveillance; Sinasc=Live Births Information System.Components Use of confidential death enquiry sheet % of FHS teams that investigate deaths at the homes % of FHS teams that investigate outpatient deaths % of PACS teams that investigate deaths at the homes % of PACS teams that investigate outpatient deaths % of NEH sending hospital death investigation occurred in their facilities % of deaths with hospital investigation reports submitted by NEH % of NEH facilitating access to medical records for investigation % of deaths with proper home investigation % of deaths with appropriate investigation at the prenatal healthcare facility % of deaths with appropriate investigation at the maternity unit where the expectant mother received care for delivery % of deaths with appropriate investigation at the facilities where the child was assisted % of deaths with appropriate investigation at the healthcare facility where the death occurred % of deaths with investigation at the necropsy services % of deaths with completed home, outpatient, hospital or necropsy service investigation % of deaths with all dimensions of the investigation report completed Adequate completion of the confidential death enquiry sheet Computer available for IMS activities in sanitary district Printer or copier available for SIM activities in the sanitary district Paper available for SIM activities in the sanitary district Phone/Fax available for SIM activities in the sanitary district Internet speed installed in sanitary district Sufficient number of death summary sheets in the sanitary district Meeting room available to discuss the death Official record (minutes) of death discussion meetings Alternative use to document death discussion meetings % of deaths with the summaries for discussion Summons of Technical -Management Group to discuss death via internal communication % of discussions with the participation of the ESF medical team % of discussions with the participation of the ESF nursing team % of discussions with the participation of the ACS of the ESF team % of discussions with the participation of the PACS nurse % of discussions with the participation of the CHA of the PACS Yes to 100%

Components%
of discussions with the participation of a representative of the outpatient unit involved in care % of discussions with the participation of a representative of the hospitals involved in maternal and child care % of discussions with the participation of a representative of the health facility where the death occurred % of discussions with the participation of health inspection managers / technicians of the sanitary district % of discussions with the participation of health inspection managers / technicians at central level % of discussions with the participation of child health policy managers / technicians of the sanitary district % of discussions with the participation of child health policy managers / technicians at central level % of discussions with the participation of women's health policy managers / technicians of the sanitary district % of discussions with the participation of women's health policy managers / technicians at central level % of discussions with the participation of primary care policy managers/technicians of the sanitary district % of discussions with the participation of primary care policy managers/technicians at central level % of discussion with the participation of the sanitary district manager Use of summary sheet to conclude the death after case discussion % of deaths with completed summary sheet % of deaths with health promotion and care measures to prevent similar identified deaths % of deaths with classified avoidability % of deaths for which changes were requested to the birth certificate/Sinasc after discussion of the Technical -Management Group % of deaths for which changes were requested to the cause of death/SIM after discussion of the Technical -Management Group % of deaths for which changes were requested to other variables of the death certificate / SIM after discussion of the Technical -Management Group % of deaths discussed in the extended Technical -Management Group % of deaths discussed in the restricted Technical -Management Group % of deaths discussed in the restricted and extended Technical -Management Group % of deaths investigated and discussed in the Technical -Management Group in a timely manner Average time between date of death and conclusion of the case discussion to Resultados em proporção 80-100%=1; 60-79%=0,75; 40-59%=0,5; 20-39%=0,25 e <20%=0.-Sim=1 ou Não=0.-Capacidade do computador: acima de 2 núcleos=1; de 1 a 2 Gigahertz (GHz) com 2 núcleos=0,75; de 1 a 2 GHz com 1 núcleo=0,5; até 1 GHz com 2 núcleos=0,25; até 1 GHz com 1 núcleo=0.-Capacidade da internet: acima de 15 Megabites por segundo (Mbps)=1; de 11 a 15 Mbps=0,75; de 6 a 10 Mbps=0,5; de 1 a 5 Mbps=0,25; menos de 1 Mbps=0.-Registro da reunião Ata=1 ou outro documento=0 DO= declarações de óbito; DN= declarações de nascidos vivos; DS= distrito sanitário; NC= nível central; SIM= Sistema de Informação sobre Mortalidade; NEH= núcleo de epidemiologia hospitalar; ESF= estratégia de saúde da família; PACS= Programa de Agentes Comunitários de Saúde; VOI= vigilância do óbito infantil; Sinasc= Sistema de Informação sobre Nascidos Vivos.

Components/Indicators ( n=29 )
Identification of infant deaths (12 indicators with weight 10) Transport available for MIS activities Technician competent to provide feedback on death certificates Processing power of computers used in MIS activities Internet speed installed at central level Printer or copier available for MIS activities Paper available for MIS activities Technician competent to sort the children's death certificates Number of technicians responsible for IMS at central level Transport available for IMS activities at central level Number of technicians responsible for IMS in sanitary district Persons responsible for IMS in the sanitary district participated in training, lectures, courses or sought to increase their awareness of this surveillance Sufficient number of specific forms to validate home addresses in the sanitary district Scores obtained Epidemiological investigation of child deaths (7 indicators with weight 30) Persons responsible for IMS at central level participated in training, lectures, courses or sought to increase their awareness of this surveillance Sufficient number of confidential record of death investigation in sanitary district Transport available for IMS activities in sanitary district % of FHS team members trained for IMS % of PACS workers trained for IMS % of NEH trained to investigate deaths at the hospital % of technicians responsible for IMS at central level with competence to investigate deaths at the necropsy services Scores obtained Discussion of infant deaths (7 indicators with weight 40) Computer available for IMS activities in sanitary district Printer or copier available for MIS activities in the sanitary district Paper available for MIS activities in the sanitary district Phone/Fax available for MIS activities in the sanitary district Internet speed installed in sanitary district Sufficient number of death summary sheets in the sanitary district Meeting room available to discuss the death Scores obtained Submission of health promotion and care proposals and correction of official statistics (3 indicators with weight 20)Person inputting data at central level is competent to apply with corrections proposed by the Technical -Management Team in the Sinasc and SIM databases % of summary sheets sent from sanitary district to central level Technician responsible for IMS at central level is competent to enter data in the summary sheet at the MIS-

Table 1
Matrix of indicators and parameters by component and approaches of infant mortality surveillance in Recife.

Table 1 continuation
Matrix of indicators and parameters by component and approaches of infant mortality surveillance in Recife.

Table 1 continuation
Matrix of indicators and parameters by component and approaches of infant mortality surveillance in Recife.
-Meeting record Minute=1 or other document=0 DC=death certificate; BC=live birth certificate; SD=sanitary district; NC=central level; MIS= Mortality Information System; HEU=hospital epidemiology unit; FHS=family health strategy; PACS=Community Health Agent Programme; IMS=infant mortality surveillance; Sinasc=Live Births Information System.

Table 2
Indicators of the infant mortality surveillance structure by component according to recorded values and judgment.Recife, 2014.DC=death certificate; BC=live birth certificate; SD=sanitary district; NC=central level; MIS= Mortality Information System; HEU=hospital epidemiology unit; FHS=family health strategy; PACS=Community Health Agent Programme; IMS=infant mortality surveillance; Sinasc=Live Births Information System.

Table 3
Indicators of infant mortality surveillance by component according to values found and judgmentRecife, 2014.