The Incidence of Deaths due to Unsafe prediction through a linear regression model for in 2017 Taís Melo Torres - [email protected] Bernardo Lanza Queiroz – [email protected] Universidade Federal de Minas Gerais The Problem

Social repression towards Underreporting of deaths Criminalization of women who went through resulting from * an abortion Abortion

The Objective

Propose a regression model to estimate the incidence of mortality due to Unsafe Abortion, correcting the underreporting.

The Solution Build a linear regression model** that relates Maternal Deaths due to Abortion (in general) to the explanatory variables Adolescent Fertility Rate, Contraceptive Prevalence and Maternal Mortality Ratio, selecting data from countries wich records aren’t affected by underreporting and use it along with Unsafe Abortion Rate to predict the incidence of deaths due to Unsafe Abortion in Brazil. ➢ Introduction Deaths due to Unsafe Abortion – all those which are not legal under the Brazilian constitution - are a public health problem in Brazil (FERREIRA, 2017)

LEGAL ABORTION* Why is Unsafe Abortion a Public Health Problem? a)If there is no other way to save 25 million Unsafe Poorer women the life of the pregnant woman and are more practiced by a doctor. occurred annually exposed to b)Pregnancy resulting from . between 2010 complications and 2014 from abortion ILLEGAL ABORTION* (OMS, 2017) Abortion is a (MENEZES; AQUINO, 2009) common procedure that Any other situations affects women from all social classes * According to Brazilian constitution. (MENEZES; AQUINO, 2009) Selected Data - Model Selected Countries

Germany Argentina Australia Austria Bulgaria Canada Chile Croatia • TMA - Maternal • TFA – Adolescent Cuba USA Finland France Death Rate due to Fertility Rate (15 to

Abortion, both 19 years old, 2017); Netherlands Hungary Ireland Italy Variable induced and Variables • PC – Contraceptive Japan Malaysia Mauritius Norway spontaneous (2017). Prevalence by any Oman Poland Portugal Czech Republic methods (mean of Romania Russia Serbia Switzerland the years 2000 to 2017); Dependent • TMT – Maternal Explanatory Mortality Ratio (2017). Explanatory Variables – Brazil Countries’ records were considered reliable when belonging to Group I in Used along with the model to predict ln(TMA) for Brazil in 2017 Luy’s Comparative Population Studies published in 2006.

•TFA – Adolescent Fertility Rate (15 to 19 years old, 2017) = 0,0591 •PC – Contraceptive Prevalence by any methods (mean of the years 2000 to 2017) = 80,25% •TMT – Maternal Mortality Ratio (2017) = 1,6098 Descriptive Analysis – Explanatory Variables

Scatter Chart and Trend Line Scatter Chart and Trend Line Dependent variable and Adolescent Fertility Rate Dependent variable and Contraceptive Prevalence 0,20 100% 0,15 80% 60%

0,10 PC TFA 40% R² = 0,5191 R² = 0,4212 0,05 20% 0,00 0% -6,00 -4,00 -2,00 0,00 2,00 -6,00 -4,00 -2,00 0,00 2,00 ln(TMA) ln(TMA)

Scatter Chart and Trend Line Residual Analysis Dependent variable and Maternal Mortality Ratio Dependent Variable 100 80

60 TMT 40 Residuals 20 R² = 0,8092 0 -6,00 -4,00 -2,00 0,00 2,00 Observations - Dependent Variable ln(TMA) Results - Model Results - Incidence of Deaths due to Unsafe ln 푇푀퐴 = −3,7005 + 22,1580 ∗ 푇퐹퐴 − 1,2048 ∗ 푃퐶 + 2,0889 ∗ 푇푀푇 Abortion, Brazil 2017 Standard Term Coefficient t p-value 퐼푛푐𝑖푑푒푛푐푒 표푓 퐷푒푎푡ℎ푠 푑푢푒 푡표 퐴푏표푟푡𝑖표푛 Error 푇푀퐴 Intercept -3,7005 0,6003 -6,1640 2,28E-06 = 퐵푅 ∗ 퐹푒푚푎푙푒 푃표푝푢푙푎푡𝑖표푛 100.000 TFA 22,1580 7,3294 3,0230 0,0059 PC -1,2048 0,8391 -1,4360 0,1640 TMT 2,0889 0,3272 6,3850 1,33E-06 퐼푛푐𝑖푑푒푛푐푒 표푓 퐷푒푎푡ℎ푠 푑푢푒 푡표 푈푛푠푎푓푒 퐴푏표푟푡𝑖표푛 = Adjusted R-squared = 0,8398 퐼푛푐𝑖푑푒푛푐푒 표푓 퐷푒푎푡ℎ푠 푑푢푒 푡표 퐴푏표푟푡𝑖표푛 ∗ 푈푛푠푎푓푒 퐴푏표푟푡𝑖표푛 푅푎푡푒

Results – Unsafe Abortion Rate Item Value Lower Limit Upper Limit

푇표푡푎푙 푈푛푠푎푓푒 퐴푏표푟푡푖표푛 Unsafe Abortion Rate = ln(TMA) 0,0052 -0,0193 0,0297 (푇표푡푎푙 푁표푡 퐼푙푙푒푔푎푙 퐴푏표푟푡푖표푛+푇표푡푎푙 푈푛푠푎푓푒 퐴푏표푟푡푖표푛)

ICD - 10 2013 - 2017 TMA 1,0052 0,9809 1,0301 O03 77 Incidence of Deaths due to O04 6 1.062 1.036 1.088 Abortion Not Illegal Abortion 83 Incidence of Deaths due to O05 47 796 777 816 O06 159 Unsafe Abortion O07 41 O08 2 Unsafe Abortion 249 The limits presented consider a confidence interval of 95%, obtained in Unsafe Abortion Rate 75,00% Fonte: MS/SVS/CGIAE - Sistema de Informações sobre mortalidade - SIM R software when the variable TMA was predicted. Results – Comparison with other Data Sources and Discussion

Incidence of Deaths due to Unsafe Abortion in Brazil, 2017 Estimated x Observed • there is a very significant underreporting of the female deaths due to Unsafe Abortion in Brazil, as Prediction - Upper Limit we expected. 816 • we estimate between 732 and 771 deaths due to Prediction - Mean Value 796 Unsafe Abortion in Brazil were not reported; Prediction - Lower Limit • The Unsafe Abortion is one of the main causes of 777

IHME maternal mortality in Brazil, even though it is 206 considered avoidable. Measuring the incidence of this cause of death is primordial for the 45 improvement of maternal health, but is only one of the many steps to attain this important objective. BENEFITS LIMITATIONS • Data selected for the model don’t need any adjustments; • The Linear Regression is not a complete description of the • Easy reproduction; relationships between variables; • Demands little operational effort; • Model built from 28 observations only. • Necessary data is freely accessible.