Epidemiological investigation of perinatal deaths in Recife-Pernambuco: a quality assessment

of perinatal deaths in a ABSTRACT Objective: to evaluate the completeness of perinatal death investigation sheets, stratifi ed by age components. Method: descriptive study carried out in Recife, PE, in 2014. Among 308 perinatal deaths, 46 were excluded from this study due to association with congenital malformations, and 7 due to missing investigation sheets. Analysis included 255 deaths (160 fetal deaths, and 95 preterm neonatal deaths). The degree of completeness of 98 variables was calculated. They were aggregated into six blocks: identifi cation, prenatal care, birth care, family characteristics, occurrence of death and conclusions and recommendations. Results: the median rate of completeness for perinatal death investigation sheets was 85.7% (82.8% for records of fetal deaths and 89.5% for records of preterm neonatal deaths). The best-fi lled information block was “identifi cation” (96.1%), as well as its components: fetal (94.7%) and preterm neonatal (97.9%). The worst was “prenatal care” (69.8%), along with its components: fetal (73.8%) and preterm neonatal (67.4%). Conclusion: investigation sheets had good completeness; there were differences between variables and components of perinatal death. Descriptors:


INTRODUCTION
Deaths in the perinatal period, which range from the 22nd week of gestation to the sixth day of life, are evidence of socioeconomic vulnerabilities and compromised quality of maternal and child healthcare (1)(2) . Worldwide, there are an estimated 4.9 million perinatal deaths per year, including 2 million fetal deaths and 2.9 million preterm neonatal deaths. The perinatal death estimate for Brazil is 53,170 (3) ; most deaths were preventable and often remain uncounted (4) .
Although the World Health Organization (WHO) launched in 2014 a plan of action to significantly reduce preventable deaths and fetal deaths by 2035 (5) , the organization's Sustainable Development Goals did not prioritize the monitoring of perinatal mortality rate, keeping it out of the public's attention (6) . However, countries that have vital statistics information systems with good coverage and reliable data can acquire knowledge on their perinatal epidemiological situation, planning interventions in favor of maternal and newborn health (7) .
Given the absence and insufficiency of vital information and better insight into perinatal deaths, death surveillance by combining maternal sociodemographic data, risk factors, healthcare histories and their nonconformities facilitates the understanding of the circumstances around the occurrence of death (5)(6) . In Brazil, despite the fact that infant and fetal death surveillance became mandatory in the health services that make up the Unified Health System (SUS) only in 2010, there were several successful death surveillance experiences prior to the constitution of this normative base (8)(9) .
Studies evaluating all stages of the death surveillance process and its investigative tools are still meager in the country (10) . Currently available analyzes on the consistency of information obtained from household interviews and institutional records, and on the completeness of investigation sheet variables, are insufficient. The same is not observed in respect to declarations of live births and deaths, which are heavily studied (11) .
Knowledge on variables presenting non-null records in perinatal death investigation sheets may contribute to improvements in terms of gathering vital statistics, investigation quality, epidemiological situation diagnosis, evaluation of maternal and child healthcare, and planning of perinatal public policies.

OBJECTIVE
To evaluate the completeness of variables in perinatal death investigation sheets, stratified by age components.

Ethical aspects
This research project was approved by the Research Ethics Committee of the Health Sciences Center of the Federal University of Pernambuco on July 1, 2015 and obtained an approval from the Health Secretariat of Recife.

Study design, place and period
This was a descriptive study on the completeness of perinatal death investigation sheets carried out in 2014 with residents of Recife, PE. Recife is the capital of the state of Pernambuco, located in the Northeast region of Brazil, which in 2014 had 1,608,488 inhabitants, distributed over 218 km 2 , 94 neighborhoods and six sanitary districts (12) . Surveillance of infant and fetal death was implemented in 2002, after the situation of infant mortality was analyzed. Death surveillance was collectively constructed by public managers and healthcare workers. Its goals were: to identify failures in maternal and child healthcare, defining their avoidability; to improve the quality of information systems and employ them in the research, planning and adoption of measures aimed at reducing infant mortality (9) .

Inclusion and exclusion criteria
In 2014, 308 perinatal deaths were recorded in the Mortality Information System (SIM), including those associated with mothers who resided in Recife. Perinatal deaths due to congenital malformations and those with missing investigation sheets were excluded.

Data source
Mortality Information System (SIM) and perinatal death investigation sheets. This sheet includes all the variables that the Ministry of Health requires to be monitored (4) , alongside others that are of local interest.
Study protocol, analysis of results and statistics Structured variables (n = 98) were analyzed in eight blocks: notification and investigation of death; identification; prenatal care; birth care; family characteristics; occurrence of death; and conclusions, recommendations and preventive measures.
Completeness was analyzed by the proportion of filled fields in each variable and each block, and presented according to fetal and preterm neonatal components. Investigation sheet data were coded and typed with double entry, for inconsistency analysis and to minimize errors, using the Epi Info software, version 7.0. In order to classify the filling level of the variables, the scale proposed by Romero and Cunha (13) was adopted: excellent (more than 95.0% filled); good (90.1 to 95.0%); regular (80.1 to 90.0%); poor (50.1 to 80.0%) and very poor (50% or less). Descriptive statistics were performed through R software, version 3.2.2.

RESULTS
Among 308 perinatal deaths, 46 (14.9%) were excluded from the study because they were associated with congenital malformation, and 7 (2.3%) due to missing investigation sheets. Analysis included 255 deaths (160 fetal deaths, and 95 preterm neonatal deaths). The variables analyzed presented a median filling level of 85.7%. For fetal and preterm neonatal deaths, the median filling level was 82.8% and 89.5%, respectively. In the perinatal general category, the best-filled block of variables was "identification," with a median filling level of 96.1%. The worst was "prenatal care," with 69.8%. In both perinatal components, the "identification" block presented the best completeness (fetal: 94.7%; preterm neonatal: 97.9%). Similarly, the block with the worst completeness among components was prenatal care (73.8% and 67.4%) ( Table 1). Table 2 presents data on death identification. Among a total of 11 variables, 7 had "excellent" completion. The variable with the lowest completion proportion (40.0% = "very poor") in perinatal death investigation sheets was "partner occupation." Among 19 variables related to prenatal care, 4 were classified as having "excellent" completion and 6 were considered as having "very poor" completion.
In the "family characteristics" block, among 12 variables, 8 were classified as having "regular" completion, and 4 as having "poor" completion. As for the fields related to "occurrence of death," among 11 variables, 3 had "excellent" completion, and 3 had "very poor" completion ("perinatal" general category). Variables with the best filling levels were "place of occurrence" (99.4%) and "necropsy" (99.4%), in the fetal component. In the preterm neonatal component, "place of occurrence" and "unit type" stood out with 100% completion (Table 4).
In the block of variables related to conclusions and recommendations, 5 (8.9%) were considered to have "excellent" completion. Variables with the best percentage of completeness in the perinatal general category were "avoidability classification," and whether the investigation changed or corrected the cause of death, with 100% completion for both (Table 5).

DISCUSSION
The completeness of the research records was classified as "good," and the comparison between components showed that almost all of the analyzed data blocks had better completeness in the preterm neonatal component. Data on stillbirths are usually deficient and of inferior quality, mainly in respect to sociodemographic and care-related information (5,14) . The low quality of data on stillbirths limits its use for the proposition of actions directed at its confrontation (2) .
The importance (or lack thereof) given to filling information on fetal mortality is one of the main factors associated with incompleteness of data (15) . In Brazil, despite the declining trend of the stillbirth rate, it is still almost twice as high the one found in developed countries (16) . In addition, mortality differentials between states are significant (17) .
Most "identification" block data were rated as "excellent" in both components, similar to what was found by recent assessments of vital statistics information systems, which demonstrated improved data quality (9,11) . After the implementation of infant and fetal death surveillance in Recife, confirmation of provided information became a requirement, and the proportion of blank and ignored fields was significantly reduced (10) .
Regarding variables in the "prenatal care" block, "risk factors," "personal antecedents," "interval between the last two pregnancies," "pre-gestational weight" and "maternal height" presented "very poor" completeness. Admittedly, data on preexisting maternal conditions are poorly reported (18) . While an interview with the mother is considered the gold standard for obtaining information on maternal behaviors (19) , the inadequate completion of this information makes it difficult to understand the circumstances of death, the evaluation of the assistance offered to the pregnant woman and to the fetus or newborn, and the evaluation of socioeconomic conditions. This data could assist in interventions to reduce the occurrence of avoidable deaths (10,13) .
A study carried out in Alagoas (11) showed that the filling level of variables related to prenatal care was poor, and a study carried out in Pernambuco (20) also pointed out a greater inadequacy of the prenatal care block in investigation sheets of a group of children who did not receive medical discharge after birth. Inadequate completion of information on prenatal care makes it difficult to assess maternal and child care, and perpetuates deficiencies mainly in primary care and in the prevention of avoidable deaths. In this sense, a study showed that mothers with inadequate prenatal care have a higher chance of death (21) , and a recent evaluation of quality of prenatal care provided in the basic health network throughout Brazil revealed that only 15% of respondents received quality care according to the following parameters: number of visits, vaccination status, prescription of ferrous sulfate, physical examination, medical advice and complementary examinations (22) .
Birth information obtained a classification of "good." A systematic review study found that procedures and conditions that occur close to birth have more reliable records, probably due to better access to clinical records that make it possible to complete the information (23) . At the national level, the Information System on Live Births (Sinasc) was evaluated as good quality and as having high completeness (24) . A hospital investigation of perinatal deaths in Ethiopia showed their avoidability, and that factors related to the health worker were the most commonly identified, followed by those referring to the patient and administrative factors (19) . Measures to improve the quality of care are linked to the improvement of information quality and completeness, in order to allow the recognition of each health service's current situation, allowing for the planning of adequate interventions.
Variables related to family characteristics were classified as "good." The evaluation of the socioeconomic context in which death occurred contributes to the planning and implementation of intersectoral actions (10,25) .
The completeness of data related to conclusions and recommendations was good, an indication that the completion of the investigation steps was satisfactory and provided an assessment of problems, as well as the suggestion of measures to prevent avoidable perinatal deaths (8,10) . The change or correction of the cause of death can be considered indicative of the adequacy of infant death surveillance actions (26) .

Limitations of this study
The study is limited by its use of a non-specific completeness score to evaluate perinatal death investigation sheets. However, the small number of studies analyzing the investigation sheet and its contribution to death surveillance attests this study's relevance.
Contributions to the area of nursing, health or public policy Complete epidemiological investigation optimizes the targeting of public resources and actions to reduce perinatal mortality. Analysis of completeness allows the evaluation of the information and contributes to its qualification. For the death investigation to be fully successful, it is essential for it to have reliable information, so that it can properly intervene on the identified gaps and propose effective measures for the prevention and reduction of perinatal mortality.

CONCLUSION
This assessment of the completeness of perinatal death investigation sheets ranked overall completion level as "good." However, there are considerable differences in completeness between variables and components. In order for death surveillance to play its role of providing information on deficiencies in the maternal and child healthcare process and help direct interventions to the issue of avoidable deaths, it is crucial to ensure the proper filling out of the investigation sheet.