Evaluation of the adequacy of information from research on infant mortality in Recife , Pernambuco , Brazil

This study is an evaluation of infant death research in Recife, Pernambuco (PE). It is a cross-sectional study with 120 variables grouped into six dimensions (prenatal, birth, child care, family characteristics, occurrence of death, and conclusion and recommendations), weighted by consensus technique. The research was classifiedas adequate, partially adequate or inadequate according to a composite indicator assessment (ICA). There was dissension on 11 variables (9 in prenatal dimension, one in labor and birth, and 1 in the conclusions and recommendations). Of the 568 deaths studied, 56.2% have adequate research. The occurrence of death was the best-evaluated dimension and prenatal the poorest. The preparation of the ICA enables professionals and managers of child health policies to identify bottlenecks in the investigation of infant deaths for better targeting of actions, and contributing to the discussion about surveillance in other cities and states.


Introduction
The infant mortality surveillance (IMS) is recommended as a strategy that contributes to the prevention of avoidable deaths by health services 1,2 .In Brazil it was adopted by the Ministry of Health (Ministério da Saúde -MS) as a state policy in 2010, with the publication of a normative and legal basis 3 , beginning with some municipalities that developed this strategy [4][5][6][7][8] .
In spite of the scarcity of evaluations of IMS in the country, studies reiterate that this strategy represents an important administrative tool, capable of providing information for healthcare teams, generating critical awareness, and aiding the improvement of information systems 5,9,10 .It also facilitates planning interventions directed at the main healthcare problems and assistance barriers 2 , awakening particular interest in reaching the target of infant mortality reduction that is a constant goal of the Sustainable Development Objectives (Objectivos de Desenvolvivemento Sustentável-ODS) that are to be reached by the signatory countries by 2030 [11][12][13] .
In the formulation of the IMS, the fatalitiesare identified, investigated, and discussed in specific forums, generating healthcare promotion proposals, as well as correction of vital statistics 5,9,10 .The death investigation retrospectively permits an evaluation via patient records and through interviews with family and professionals involved in maternal and child care 10,14 .In order to unravel the circumstances in which the event occurred, the investigation explores the variety of determinants in infant mortality 12 , identifying failures in the chain of healthcare actions 6,[15][16][17][18][19] .
The instrument of data collection regarding infant death is not the same in all of Brazil, in spite of the Ministry of Health making a standard model available for investigation in the domestic, outpatient, and hospital environments 1,10 .For the municipalities that opted for their own instruments, collection of a group of variables that are monitored by the Ministry is obligatory 3 .
The few studies on the investigation of infant fatalities do not consider the importance and adequacy of the information toward the clarification of each case's singularity 6,8,10 .The incorporation of these elements in research with an evaluative approach is indispensable for elucidating the limitations and possibilities of the investigation of infant death, with an eye towards its full implementation in various states and municipalities in the country.
In this way, this study has the objective of evaluating the adequacy of investigations into infant deaths in Recife (PE) occurring between 2011 and 2013.

Methods
This is a cross-sectional study with an evaluative framework about the IMS carried out in Recife (PE), a capital city in northeast Brazil, which has a totally urbanized area of 218.5 km 2 , with 94 neighborhoods distributed in six health districts.In 2010, the population was 1,537,704 inhabitants, with 19,142 (1.24%) being children less than one year old 20 .
Since 2006, all infant deaths of mothers residing in the municipality were investigated, with the exception of birth defects.The IMS is comprised of identification of infant deaths; epidemiological investigation; discussion of the deaths; recommendation of proposals for health promotion, healthcare, and correction of vital statistics; developed by a multidisciplinary team, decentralized by health districts.The investigations are conducted by different professionals: clinically by the Centers for Clinical Epidemiology, autopsy service by the core team, and outpatient and house-call services by workers in primary care.
A proprietary confidential form was used, which considers the variables recommended by the Ministry of Health 1,3 related to: identification of the child and the mother; characteristics of the family; prenatal data; childbirth and delivery; childcare; the occurrence of the death and the conclusions and recommendations.Professionals in healthcare assistance, surveillance, and administration participate in the discussion, distinguishing the process from other initiatives that occur in the context of the committees 4,5 .The recommendations are forwarded to the health administrators for the appropriate actions.
The research was carried out with records of infant deaths that occurred in the years 2011 to 2013, in order to work with recent data up to the last year with a complete database.In this period, of the 628 deaths eligible for investigation, 568 (90.4%) had confidential investigation records located and made available by the Secretary of Health of Recife, constituting the study's population.
Considering the differences in the factors that contributed to the death occurrence, independent of age range, the infant fatalities were grouped into the following categories: deaths of children that did not receive a hospital discharge after birth (Group 1), and deaths of children that received a hospital discharge to the home after birth (Group 2).In the period studied, all of the children who died were born in a hospital environment.Group 1 included the majority of neonatal deaths and part of the post-neonatal deaths, with causes originating in the perinatal period; Group 2 included the majority of post-neonatal deaths, with causes originating in this period.
Being indispensable to the beginning of the investigation, all of the variables of the confidential investigation record of the infant fatalities were investigated, except those related to the identification of the death, as well as those variables that were repeated in other aspects.For groups 1 and 2, 106 of 120 variables were studied, distributed into five dimensions: pre-natal, pregnancy and childbirth; family characteristics; occurrence of death; and conclusions and recommendations.Childcare was added to Group 2.
The selected variables were evaluated according to their importance in the investigation of infant death by the technique of consensus 21 , by a group of 25 specialists in the areas of maternal-infant health and death surveillance, working in the different levels of the Unified Health System (Sistema Único de Saúde -SUS).Initially, a meeting was held with the specialists for a knowledge briefing on the study's proposal and the criteria for evaluation of the variables: value of the information for the reconstruction of the history of each death; capacity to identify the level of the system and the action in the chain of healthcare in which possible failure occurred; possibility of understanding the events that contributed to the death event; and to generate recommendations for its avoidance.For each variable, a weighted value between 0 and 10 must be a given.
A classification framework for analysis was mailed electronically to each of the specialists and only one did not respond.Using the established weights, the average (x)and standard deviation (σ) for each variable were calculated.The average indicated the importance of the variable from the point of view of the specialists, and the standard deviation indicated the degree of consensus.Thus, σ ≤ 1 was considered consensus, σ > 1 and < 3 little consensus, and σ ≥ 3 dissension.Chart 1 presents a synthesis of the stages of the technique of consensus.
All of the variables in Group 1 achieve consensus or little consensus.In Group 2, dissension in 11 variables was identified (nine related to the pre-natal dimension, one to pregnancy and childbirth, and one to conclusions and recommendations), which were discussed in person by 19 specialists for consensus according to the new weight (Table 1) Simultaneously, the variables of the investigated death records of the period under study were evaluated by the researchers regarding their degree of completion, attributing the following weighted values: 0 (not filled out), 1 (partially filled out), and 2 (filled out).For each variable of the death investigation, the value observed (weighted value of importance x completion) and the maximum desired value (weighted value of importance x maximum completion value) were obtained.
Next, a composite indicator assessment (indicador composto para avaliação-ICA) for each dimension (ICA dim ) and for the set of dimensions (ICA total ) of infant death investigations was constructed.The ICA dim represented a proportion of the sum of observed values of the set of variables of the respective dimension in relation to the sum of maximum desired values for each of the same variables, according to the following equation: To obtain the ICA total of the investigation, the average of the ICA dim was used.Using the values of the compound indicators, the investigation of each infant death by dimension and by its set were classified as: adequate (ICA ≥ 80%), partially adequate (ICA = 60-79%), and inadequate (ICA < 60%) (Table 1).The differences in the investigation between the health districts and the dimensions were verified by the Pearson chisquared test (χ 2 ), with a significance level of 5%.When more than 20% of the cells presented a desired value < 5, this test was not applied.
The study was approved by the Research Ethics Committee of the Aggeu Magalhães Research Center of the Oswaldo Cruz Foundation -CPqAM/Fiocruz and obtained permission from the Secretary of Health of Recife.

Results
In the period studied, 843 infant deaths occurred in Recife, with an infant mortality coefficient (coeficiente de mortalidade infantile -CMI) of 12.4 per 1000 live births (nascidos vivos -NV).The largest number of infant deaths (198) was recorded in the health district VI, while the greatest CMI (14.7 for 1000 NV) was found in district I.Of the 628 deaths eligible for investigation, 96.7% (607) were investigated and discussed.In the districts, this proportion varied from 92.9% (III) to 100.0% (I and V).Of the 568 deaths studied, 467 (82.2%) belong to Group 1, and 101 (17.8%) to Group 2 (Table 2).

Ʃobserved value Ʃmaximum desired value
In Recife, 56.5% of the investigations in Group 1 were classified as adequate.Differences between the dimensions of the investigation were verified, whose adequacy varied from 56.7% (pre-natal) to 81.2% (occurrence of the death).The best evalua-tion of the investigation (87.8% adequate) was in health district VI, and the greatest percentage of inadequacy was in III (41.4%) (Table 3).
In Group 2, 54.5% of the investigations were adequate, with the dimensions of family char-Chart 1. Synthesis of procedures for the creation of compost indicators for evaluation of investigation of infant deaths.

Phase Activity developed Actors Grouping of infant deaths
Death records were made available in: Group 1: deaths of children that did not receive a hospital discharge after death.Group 2: deaths of children that received a hospital discharge to the home after birth.

Selection of variables of IMS instruments
Evaluation of all variables of the investigation records of infant deaths, with exceptions for those related to the identification of death indispensable to the beginning of the investigation, and those that are repeated in other categories, totaling: Group 1 = 106 variables distributed in five categories (pre-natal, birth and delivery, family characteristics, occurrence of death, and conclusions and recommendations.Group 2 = 120 variables distributed in six categories (same as Group 1 plus childcare)

Researchers
System for agreement study's proposal and criteria for evaluation of the variables.
2. Electronic mailing of the relevant guidelines for analysis of each variable, to be given a weight between 0 and 10.
3. Using the established weights, the average and standard deviation for each variable were calculated.In the health districts, the adequacy of the investigation varied from 21.1% (III) to 85.7% (VI).Between the districts there were also differences The investigation validated or altered one or more of the filled-out variables of the DNV or the DO?If so, which?
The investigation validated/altered the sequence of causes of death?If so, describe the causes of death after the investigation 9,4 The discussion of the death generated specific recommendations for the inherent problems identified at the health sector 9,5 1,0 9,5 1,0 NA = not applicable.*Variables that had dissension and were re-discussed by the specialists.Group 1: deaths of children that did not receive a hospital discharge after birth.Group 2: deaths of children that received a hospital discharge for the home after birth.the most adequate (80.8%) and pre-natal being the most inadequate (30.6%).Health district VI was the best evaluated, with 87.3% of investigations classified as adequate, and district III was the worst (40.7% inadequate) (Table 4).
In Groups 1 and 2 and in the total of infant deaths, a significant difference (p < 0.001) was observed between the health districts and the dimensions (pregnancy and birth, occurrence of death, conclusion and recommendations, and the set of dimensions) in which it was possible to calculate χ 2 (Tables 2 and 3).are considered priorities and with the obligatory operationalization by decree of the Ministry of Health 3 .
National and international research pertaining to the auditing and investigation of infant death mainly deals with the circumstances of the death occurrences, the quantitative aspects of the investigated fatalities, and the state of completion of the records 5,6,8,10,14,16,18,19,22 .In the present study, in addition to these approaches, we incorporate other elements that permit us to construct a composite indicator for evaluation of the infant death investigation.To evaluate the investigation records and the data pertaining to the closing of the case after discussion, a large number of vari-ables of the instrument were used.The method utilized made possible an evaluation of the deaths' investigations with greater discrimination regarding the positive points and those that need to be improved in the operationalization of this strategy, which can be employed in other municipalities.
The weights attributed to the variables by the technique of consensus represent only one degree of provisionary consensus, considering that the epidemiological profile and the healthcare assistance network are in constant change, however it functions as a point of departure and reference for a process that can include other strategies of redefinition of the variables' weights and for the renovation of consensus 21 .Nevertheless, it constitutes a useful tool for evaluation, and can be used in whole or in part, or modified, depending on state and municipal realities.
No variable from the investigation records, whether in deaths of Group 1 or 2, was excluded by the technique of consensus.This fact is probably related to the process of discussion of this strategy in the country, and in particular in the locale of the study, especially after the multiple institutional productions such as manuals and technical notes, in the context of the Millennium Development Goals (MDGs) and Sustainable Development Goals (SDGs) 1,3,[11][12][13] .
Considering the diversity of the epidemiological profile and assistance available in Brazil, the Ministry of Health suggests minimum variables that should be part of the investigation records, and recommends the addition of others that respond to local needs 1,3 .Aside from the ones proposed by the Ministry, the case under consideration adds other variables, for example some that favor a better specification of the basic cause of death by maternal affections.
The classification of infant fatalities into two groups was due to the principal activities involved in the infant death and the preferred locus of the investigation.For Group 1, the subjects of most relevance relate to prenatal care, assistance to the delivery, and the immediate care of the baby, while those in Group 2 concern the monitoring of the child.The division of fatalities into neonatal and post-natal does not entirely convey the essence of the main causes for intervention, in particular when there is use of intensive medical technologies that can delay death in newborns 17,23 .
Post-neonatal fatalities in children that did not have a hospital discharge after birth became more frequent in maternity wards that have done away with intensive therapy units.On the other hand, the improvement of life conditions of the population and access to goods and services contributed to the reduction of late neonatal mortalities among those that had a discharge from a maternity ward 17,23 .
Regarding the evaluation of the investigation by dimension, it can be stated that the occurrence of the death, pertaining to the assistance given to the child in the hospital [4][5]8,10 , presented the most adequacy for both the Groups for the total of fatalities. Sucha fact can be due to the proximity and ease of access to the medical charts by the teams of the Center of Hospital Epidemiology, which are responsible for the investigation at this level of attention, developing evaluations and institutional diagnostics 24 .
The study identified the most inadequacy of the investigation in the prenatal and childcare dimensions of Groups 1 and 2, respectively.In Group1, the prenatal data is information that is essential for the prediction of risk in the mother-child binary.In Group 2, the data on childcare is fundamental for the evaluation of growth and development and background of the child 9,25 .These are important dimensions that guide the promotion of health and well-being, as well as creating an opportunity for the identification of problems and their respective treatments since the conception of the child 4 .
The Family Health Strategy teams maintain a relevant role in the investigation in the prenatal and childcare dimensions, in that these professionals have broad access to the outpatient records.In this way, primary care is the privileged site for the collection of information about infant fatalities and activity in its determining factors, minimizing adverse conditions regarding family, territory, access to the healthcare system, and the quality of attention offered 1,6,9 .
The inadequate investigation of the prenatal and childcare dimensions, shown in this study, with a consequent lack of evaluation of the stages of mother-child care, contributed to the inequities in the investments in the various levels of attention, perpetuating deficiencies mainly in primary care.The intensification of training of the Family Health Strategy teams regarding the investigation of infant fatalities is an indispensable condition in order for the managers to know and act on the problems related to the offered healthcare services.
The low adequacy of the investigation of infant fatalities reflects difficulties in obtaining the necessary information for each case, such as: lack of access to patient records due to bureaucratic or administrative problems, principally in supplemental and private healthcare assistance services; lack of recordkeeping in the outpatient and hospital records; and illegible annotations.Regarding household interviews, the absence of information is due to wrong addresses or families moving, as well as refusals 15,22,26 .
The evaluation of the investigation of infant fatalities permitted the identification of significant differences between the health districts in their different dimensions.Evidence of deficiencies in the process of investigation makes possible the qualification and organization of the activities of the IMS in accordance with the particularities of each territory.It is noteworthy that the district, as the unit of analysis, presents the advantage of being less heterogeneous than the whole municipality.Aside from this, these territorial units are utilized for all the areas of the municipal public administration, which facilitates an intersectoral approach, planning of actions, and operationalization of social and health programs 27 .
The international experiences indicate whether the strategy is to be properly executed to understand the circumstances of the death and carry out the effective recommendations to avoid similar events [28][29][30] .Like what occurs in the improvement of information systems in health 31,32 , for the perfecting of the investigation of infant fatalities, the following should be considered: inequalities in the training of professionals responsible for this action; the lesser importance given to specific variables during the process of data collection; and the poor completion and filling out of outpatient and hospital records 25,33,34 .
For the IMS, we recommend the institution of permanent activities of training, supervision, accompaniment, and quality control of the investigations of infant fatalities for all the levels of the healthcare system involved in this strategy.The challenge of improving the investigation is a fundamental stage for the IMS to favor a critical reflexive environment concerning the care of mother and child, and contribute to the organization of thematic networks of care in the different health regions.
The development of the tool for the evaluation of IMS makes possible the identification of the bottlenecks in the investigation of infant fatalities for better targeting of actions by professionals and administrators of child healthcare policies.In addition, it contributes to reflection on this surveillance in other municipalities and states.

Collaborations
CM Oliveira MJB Guimarães and CV Bonfim worked on conception, designing, analyzing and interpreting the data, writing the article and approving the version to be published; PG Frias worked in the critical review of the article and approval of the version to be published; VCS Antonino worked in the analysis and interpretation of the data and writing of the article; ALS Guimarães worked in the analysis and interpretation of the data and ZM Medeiros worked in the critical review of the article and approval of the version to be published.

4 .
Identification of variables with dissension(σ>3) 5. Face-to-face discussion for consensus regarding a new weight for the variables with dissension (n=11 all from Group 2filling out of the record's variables Evaluation of variables from investigated death records according to their state of completion, with the following weighted values: 0 = not filled out 1 = partially filled out 2 = filled out Researchers Creation of the ICA (composite indicator assessment) 1.For each variable of the death investigation, the value observed (weighted value of importance x completion) and the maximum desired value (weighted value of importance x maximum completion value) were obtained.2. Creation of the ICA dim and the ICAtotal ICA dim = a proportion of the sum of observed values of the set of variables of the respective dimension in relation to the sum of maximum desired values for each of the same variables ICA total = average of the ICA dim Researchers Classification of the investigations The investigation of each infant death by dimension and for its set was classified as: Adequate: ICA ≥ 80% Partially adequate: ICA = 60-79% Inadequate: ICA < 60% Researchers acteristics (78.2%) and occurrence of the death (79.2%) with the best evaluation, and childcare being the most inadequate dimension (28.7%).

Table 1 .
Synthesis of consensus, according to opinion of the specialists, related to weight of variables of the investigation record of infant death.

Table 1 .
continuation it continues observed in the six dimensions, with emphasis on childcare with 46.7% of inadequate investigations in Group II (Table3).For the total of infant deaths, considering the set of dimensions, 56.2% of the investigations were adequate, with occurrence of death being

Table 1 . continuation Table 3 .
Distribution of infant fatalities in Groups 1 and 2 by dimension and health district of residence according to classification of the study.Recife, 2011 and 2013.

Table 2 .
Characteristics of the study population.

Table 4 .
Distribution of infant fatalities by dimension and health district of the residence, according to the classification of the investigation.Recife, 2011 to 2013.