IF: 0.644
REUTERS THOMSON

Predicting Unwanted Pregnancies among Multiparous Mothers in Khorramabad, Iran

AUTHORS

Farzad Ebrahimzadeh 1 , 2 , Ali Azarbar 3 , Mohammad Almasian 4 , Katayoun Bakhteyar 5 , Nasim Vahabi 2 , *

1 Instructor of Biostatistics, Department of Public Health, Faculty of Health and Nutrition, Lorestan University of Medical Sciences, Khorramabad, IR Iran

2 Department of Biostatistics, Faculty of Medicine, Tarbiat Modarres University, Tehran, IR Iran

3 Department of Statistics, Faculty of Sciences, Amirkabir University of Technology, Tehran, IR Iran

4 Instructor of English Language Teaching, Department of English language, Faculty of Medicine, Lorestan University of Medical Sciences, Khorramabad, IR Iran

5 Instructor of Midwifery , Department of Public Health, Faculty of Health and Nutrition, Lorestan University of Medical Sciences, Khorramabad, IR Iran

How to Cite: Ebrahimzadeh F, Azarbar A, Almasian M, Bakhteyar K, Vahabi N. Predicting Unwanted Pregnancies among Multiparous Mothers in Khorramabad, Iran, Iran Red Crescent Med J. 2016 ; 18(12):e26994. doi: 10.5812/ircmj.26994.

ARTICLE INFORMATION

Iranian Red Crescent Medical Journal: 18 (12); e26994
Published Online: August 24, 2016
Article Type: Research Article
Received: January 26, 2015
Accepted: March 16, 2015
Crossmark

Crossmark

CHEKING

READ FULL TEXT
Abstract

Background: Unwanted pregnancy is the kind of pregnancy which is undesirable for at least one of the parents, and is accompanied by unfavorable consequences for the family and society.

Objectives: In this study, three classification models have been used to predict the occurrence of unwanted pregnancies in the urban population in Khorramabad, Iran, and the performance of these models was compared.

Methods: In this cross-sectional study, 467 multiparous mothers referred to the health centers of Khorramabad in 2012 were selected using a combination of cluster and stratified sampling, and the relevant variables were measured. The logistic regression, decision tree, and a neural network were implemented using SPSS version 21 and MATLAB version R2013a. To compare these models, the indices of sensitivity and specificity, the area under the ROC curve, and the correct percentage of the predictions were used.

Results: Overall, the prevalence of unwanted pregnancies was 32.3%. The performance of the models based on the area under the ROC curve as the indicator was as follows: artificial neural networks (0.741), decision tree (0.731), and logistic regression (0.712). The highest sensitivity level belonged to the decision tree (73.5%), and the highest specificity level belonged to the artificial neural network (62.3%).

Conclusions: Given the high prevalence of unwanted pregnancies in Khorramabad, Iran, it is necessary to revise and improve the family planning projects. In selecting the best classification method, if the researcher is interested in the better interpretability of the results, the use of the decision tree and logistic regression is recommended; however, if the researcher is interested in a higher prediction power of the model, the neural network is recommended.

Keywords

Artificial Neural Networks Decision Tree Logistic Regression Multiparous Pregnant Mother Unwanted Pregnancy

Copyright © 2016, Iranian Red Crescent Medical Journal. This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/) which permits copy and redistribute the material just in noncommercial usages, provided the original work is properly cited.

1. Background

Unwanted or unplanned pregnancy is a kind of pregnancy which is undesirable to at least one of the parents (1, 2). Every year, about 200 million pregnancies occur worldwide, among which one-third (75 million cases) are unwanted pregnancies. Moreover, about 50 million unwanted pregnancies end in abortion every year, among which 20 million cases of abortion happen under unsafe conditions (3-5). Based on previous studies, the prevalence of unwanted pregnancies in the USA, Japan, and Tanzania is 48%, 46.2%, and 23.7%, respectively (6-8). Studies conducted in Iran have also indicated that, despite the easy availability of contraceptives, 400 thousand to 500 thousand unwanted pregnancies occur every year, among which 19% (80,000 cases) end in abortion annually (9-11). According to the statistics provided by the Iranian Ministry of Health, one pregnancy in every four pregnancies is unwanted (5, 12). Furthermore, in a systematic review conducted in Iran, the prevalence of unwanted pregnancies was about 29.7%, while another review put this figure at 30.6% (1, 2).

The most important reasons for unwanted pregnancy are the failure to use contraceptives, despite the lack of desire to become pregnant, and the improper use of contraceptives and their failure as a result (4, 9, 10, 12). In fact, social sanctions, the inequality of men and women in terms of social rights, disagreement among spouses, unavailability of modern contraceptives, inadequate planning, the inefficacy of counselling, and lack of adequate skills among healthcare employees are among the factors that make the recommended methods incompatible with the conditions of the people involved. This leads to the failure to use these contraceptive methods or the failure of the methods themselves (4).

It is evident that unwanted pregnancies per se are accompanied by undesirable consequences for the mother, baby, parents, and society, the most common of which includes illegal and unsafe attempts at abortion, which is one of the main reasons for mortality and disability among mothers. Additionally, due to the unpleasant emotions and feelings that the mother experiences during pregnancy, she faces increased stress and nervous strain and her referrals to healthcare providers are reduced considerably, which can lead to depression, suicide, and a reduction in the quality of life of the mother (3, 10, 13). Moreover, the most important effects of the unwanted pregnancy on the baby include the birth of preterm and underweight infants, anorexia, hyperactivity, harassment, and receiving inadequate attention from the parents, as well as increased infant mortality rates (3, 5, 12, 14).

Given the importance of this issue, and knowing that the occurrence of unwanted pregnancies in each society is affected by various individual and social factors, it is necessary that the determining factors in each society be studied comprehensively. Using methods known as classification models provides the opportunity to classify and predict the risk of unwanted pregnancies among these people, based on a set of variables related to the couples. The most common classification models used in such cases include: logistic regression, a decision tree, and the artificial neural network. The last two methods are considered to be among the non-parametric methods of data mining, and these methods differ from each other in terms of the classification accuracy, interpretability of the results, calculation time, and availability of statistical software applications, irrespective of the differences in the estimation methods and calculation algorithms as mentioned in many studies (15-17).

A review of the studies conducted on unwanted pregnancies shows that, in most cases, only the prevalence of this phenomenon is studied, and at most the relationship between an independent variable with the occurrence of unwanted pregnancies has been studied using t-tests and chi square tests (3, 5, 10, 12, 13, 18-24). In a limited number of studies, there have been attempts to investigate the simultaneous effects of a group of variables on unwanted pregnancies using classification models. Of course, due to the limitations related to the selection of the statistical population, the results of these studies cannot be generalized to other societies (3, 4, 9, 11, 25). The statistical models used in previous studies include simple logistic regression (3, 4, 6, 8, 11, 25-27), multinomial logistic regression9,30, artificial neural networks (25, 28, 29), and log-linear models (30). It is necessary to mention that, despite the fact that the decision tree method has been widely used in the prediction of medical outcomes (31-37), this technique has been used in few studies on unwanted pregnancies.

Given the importance of unwanted pregnancies, and knowing that up until now no comprehensive comparative study has been conducted in western Iran to predict unwanted pregnancies.

2. Objectives

The present study attempts to use three statistical models (logistic regression, decision tree, and artificial neural network) to determine the best classification method in the prediction of unwanted pregnancies in the urban population of Khorramabad, Iran.

3. Methods

3.1. Data

The population of this cross-sectional study included all pregnant multiparous mothers referred to healthcare centers in Khorramabad in 2012 to receive prenatal care. A sample consisting of 467 pregnant women was selected using a combination of stratified and cluster random sampling. To obtain the sample, first, the healthcare centers in Khorramabad were divided into three strata of the northern part, central part, and southern part of the city. Next, two centers (cluster heads) from the north, two from the south, and three from the center were selected randomly. Then, the mothers who were referred to the selected healthcare centers to receive prenatal care were entered into the sample in the order of their referral times.

The data gathering instrument used in the present study was a self-designed data questionnaire and form which the mothers completed in a self-reporting manner. For illiterate or semi-literate mothers, the form was filled out via interviews. This form included the demographic information of the mothers: name of the healthcare center, age of the couples, age difference between the couples, parity, pregnancy spacing, number of live male children, contraceptives used, educational level attained by the couples, income of the household, couples’ occupations, area of the residential unit, and possession of a vehicle. The form also included a question about whether the pregnancy was wanted or unwanted by (at least one of) the parents as the main outcome being studied.

3.2. Statistical Methods

3.2.1. Logistic Regression

This is a regression model used to analyze binary response variables, and is in fact a member of the generalized linear models (GLM), using the logit function as the link function. The general formula for this model is as follows:

Equation 1.

where α is the intercept of the model, xi is the ith independent or independent variable, and βi is the coefficient of the ith independent variable. The logistic regression can be used to predict and estimate the coefficients and effects of each independent variable. It can also be used for classification and recognition purposes.

3.2.2. Artificial Neural Network

This is a simulation of the human brain created by modelling the neurons, in which each neuron works as a processing unit. The multi-layer perceptron (MLP) neural network is one of the most widely used types of networks, and its structure includes several layers (input, hidden, and output layers). In each layer a number of activity nodes and functions is defined. The output of each layer is calculated using the sum of the weighted coefficients in that layer, and is sent to the next layer via an activity function. There are various methods and algorithms for finding the weights, but in the MLP network, the back propagation (BP) algorithm is used. Additionally, the activity function in the neural network models is similar to the link functions in the generalized linear models. As examples of the activity functions, the sigmoid and the hyperbolic tangent functions can be mentioned. In a general and simple case in an MLP with one hidden layer, the output value of the ith unit can be calculated as follows:

Equation 2.

where n is the number of observations, p is the number of nodes in the hidden layer, m is the number of nodes in the input layer (the number of independent variables), wkj is the weight of the input xij in the kth node, wk is the weight of the kth node, and β and β are the bias values of the output and hidden layers, respectively. In addition, f1 and f2 are the activity functions of the hidden and output layers of the network, respectively.

3.2.3. Decision Tree

The decision tree consists of three main elements including the root, internal node, and leaf. The normal procedure is that an independent variable is first selected as the root, and divided into several internal nodes. Each internal node is like the root divided into other nodes, until, finally, one level of the independent variable is attributed to each node. These nodes are called leaves. Usually, functions called impurity functions and an index called Gini are used to select important variables from the tree classification model. First, the value of the impurity function is calculated for the dependent variable generally based on the impurity function and the Gini index. Next, for all of the independent variables, considering the best binary division of the dependent variable, the value of the impurity function is calculated in each of the created subsets, and then their weighted averages are subtracted from the total impurity function. From among the independent variables, the variable with the highest value for this measure is selected for tree classification in the first step. In dealing with the quantitative independent variables, the best cut-off point is used for the binary division, and in dealing with the qualitative variables, each variable category is considered as a sub-branch of the classification tree. In a certain model of the decision tree called the CART, using a method called cost-complexity, the proper size (depth) for the tree is determined. This method can strike a balance between the accuracy of the decision tree and its size. Using this method, based on the classification error and the number of nodes, it can be determined which node of the tree must be pruned (16, 37).

To describe and model the data, and compare the logistic regression, decision tree, and the artificial neural network models, the software applications SPSS version 20 and MATLAB version R2008a were used. In order to select the studied variables, the logistic regression with the forward selection method was used. The significance levels for the entry and removal of the variables were P = 0.15 and P = 0.10, respectively. The probability values lower than 0.05 (P < 0.05) were considered as the statistical significance level. To fit the artificial neural network, the data were divided into two parts: network training and network testing. The network training data consisted of about 70% (327 observations) of the study data. Finally, the selected artificial neural network consisted of 10 input nodes, 7 middle nodes, and 2 output nodes, and the BP algorithm was selected. To draw the decision tree, the CART classification tree was used, and to select the important variables and their cut-off points, the impurity function and Gini index were used, respectively. Ultimately, to compare the models, the indices of the sensitivity and specificity, the area under the ROC curve, and the correct percentage of the predictions were used.

4. Results

The mean (± SD) age of the 467 multiparous mothers evaluated in this cross-sectional study was 29.7 (± 4.94) years old, with a range of 16 to 52 years old. About 74% (347) of the mothers were younger than 35 years old, and 29% (136) of them had a live male child. From among the pregnant women, 47.7% (223) had been pregnant twice, 37.1% (173) three times, and 15.2% (71) had been pregnant four times or more. The pregnancy spacing was less than 2 years in 21.2% (99) of the cases, between 2 to 4 years in 23.8% (111) of the cases, and more than four years in 55% (257) of the cases. About 80% (374) had an academic education and 74.4% (347) had an income of less than 200 USD per month. Moreover, 32% (151) of all of the pregnancies were unwanted, and 92.5% (432) of the couples who had unwanted pregnancies used contraceptives before the pregnancies. To demonstrate the relationship between the independent variables and unwanted pregnancy, the univariate analysis was used, and the results are presented in Table 1.

Table 1. Frequency Distribution of the Total and Unwanted Pregnancies Among Multiparous Mothers Referred to the Healthcare Centers of Khorramabad in 2012
VariableTotal Pregnancy, No. (%)Unwanted Pregnancy, No. (%)P Value
Mother’s age (years)< 0.001
< 35384 (82.2)112 (74.2)
≤ 3583 (17.8)39 (25.8)
Having live male child< 0.001
Yes195 (41.8)107 (70.9)
No272 (58.2)44 (29.1)
Parity< 0.001
2314 (67.2)72 (47.7)
3107 (22.9)56 (37.1)
≤ 446 (9.9)23 (15.2)
Pregnancy spacing (years)< 0.001
< 262 (13.3)32 (21.2)
2 - 4104 (22.3)36 (23.8)
≤ 4301 (64.5)83 (55.0)
Mother’s educational level0.449
Academic70 (15.0)19 (12.8)
Non-academic397 (85.0)132 (87.2)
Income (USD)0.228
< 200329 (70.4)104 (68.9)
≤ 200138(29.6)47(31.1)

Given the results of the univariate analysis, the mother’s educational level (P = 0.449) and income (P = 0.228) had no significant relationship with the unwanted pregnancy rates. However, variables such as the mother’s age, having a live male child, parity, and pregnancy spacing had significant relationships with unwanted pregnancies, and these variables were entered into the logistic regression, artificial neural network, and decision tree models as independent variables.

4.1. Findings of the Logistic Regression

The results of the logistic regression are presented in Table 2. From among the independent variables, variables such as the parity, pregnancy spacing, and having a live male child had significant relationships with the unwanted pregnancies, while only the mother's age had no significant relationship (P = 0.102). In addition, given the values of the odds ratio, the parity (3rd) with OR = 3.39 had the highest impact on the occurrence of unwanted pregnancies. Moreover, the area under the ROC curve (AUC) for this model was 0.712, and the correct percentage of the prediction was 0.638.

Table 2. The effects of the independent variables on unwanted pregnancy among multiparous mothers referred to the healthcare centers of Khorramabad in 2012
VariableOR95% CIP Value
LowerUpper
Mother’s age (y)0.102
< 35Ref---
≥ 351.6010.9102.8170.102
Having live male child0.010
NoRef---
Yes1.7911.1482.7940.010
Parity< 0.001
2Ref---
33.3922.0425.635< 0.001
≥ 42.8091.4035.6260.004
Pregnancy Spacing (y)< 0.001
< 2Ref---
2 - 40.4600.2330.9090.026
≥ 40.2570.1400.473< 0.001

4.2. Findings of the Artificial Neural Network

To make a three-layer perceptron artificial neural network model, given the training data set and the independent input variables, all of the networks based on 5 to 20 nodes in the hidden layers with a momentum of 0.80 to 0.95 and a learning rate of 0.01 to 0.40, the sigmoid and hyperbolic tangent link functions were evaluated. After examining all the possible models to match the structure of a three-layered neural network, ultimately, a network with 10 input nodes, 7 hidden nodes, 2 output nodes, and a learning rate of 0.05, with hyperbolic tangent and sigmoid functions, and a BP algorithm was selected as the best neural network. The area under the ROC curve was 0.741 and the correct percentage of the prediction was 0.650.

4.3. Findings of the Decision Tree

Given the impurity function and the Gini index, parity was selected as the input variable (the most important predictor variable), and a CART decision tree with a depth of 4 was obtained, which can be seen in Figure 1. The area under the ROC curve in this model was 0.731 and the correct percentage of the prediction was 0.648. According to this tree, the variable of the pregnancy spacing among those mothers who have had two pregnancies and those having a live male child among the mothers having had more than two pregnancies were highly important. Additionally, the mother’s age was of considerable importance in those mothers with more than two pregnancies and a live male child.

The decision tree of the unwanted pregnancies among multiparous mothers referred to the healthcare centers of Khorramabad in 2012.
Figure 1. The decision tree of the unwanted pregnancies among multiparous mothers referred to the healthcare centers of Khorramabad in 2012.

Finally, to compare the three methods, the values of the indices of sensitivity and specificity, the area under the ROC curve, and the correct percentage of the prediction were used, and the results are shown in Table 3. Based on the criteria, the artificial neural network and the decision tree were, in general, better and more able to classify and predict unwanted pregnancies.

Table 3. The values of the sensitivity, specificity, area under the ROC curve, and the correct percentage of the predictions for comparing the logistic regression, decision tree, and neural network among the multiparous mothers referred to the healthcare centers of Khorramabad in 2012
MethodAUCaSensitivity (%)Specificity (%)Correct Percentage (%)
Logistic regression0.71270.260.163.8
Decision tree0.73173.560.864.8
Neural networks0.74170.962.365.0

aarea under the curve.

5. Discussion

The present study attempted to predict unwanted pregnancies in Khorramabad, Iran, based on variables such as the mother's age, parity, pregnancy spacing, and having a live male child using logistic regression, the decision tree, and an artificial neural network. Based on the area under the ROC curve, the best methods in order of superiority were the artificial neural network, decision tree, and logistic regression.

The reason for the superiority of the artificial neural network in predicting wanted and unwanted pregnancies must be found in the nature of the learning process, because in this process, all of the complex nonlinear relationships and the interactions between the independent variables are learned; therefore, the predictive power of this model will be considerably high40. However, if the proper architecture is not selected for the neural networks, and especially if a large number of hidden layers and nodes are selected, the network may learn the random errors in the learned data (overlearning) and, consequently, despite the high accuracy of the neural networks, this method may perform poorly in the prediction of the tested data. The most important weaknesses of the artificial neural network as a black box method are the uninterpretability of the weights, complexity of the method, calculation algorithms, and calculation time; therefore, for its proper implementation, it is necessary to use statistical applications related to data mining (15-17, 38).

The decision tree, as a nonparametric method and a powerful instrument in data mining, has shown high accuracy in the prediction of unwanted pregnancies as well. The results of the simulation studies showed that when the data are highly stretched and skewed, or when a large percentage of the variables are qualitative (as in the present research), the decision tree is a good technique for data mining18. The most important advantage of this method is its considerable interpretability due to its tree structure; therefore, when one is more interested in the interpretability of the results rather than the high prediction accuracy, this is one of the best options (19,34-39). However, one of the problems of the decision tree is that it decides based on just one variable in each stage of the algorithm and does the division process (37).

The most important advantage of the logistic regression method is that it, as a model-based method, is capable of testing and comparing the impact of each of the independent variables on the occurrence of unwanted pregnancies. As a result, it offers desirable interpretability via a model which is presented in the format of a closed form. For instance, in the present study, it can be stated that a pregnancy spacing of less than two years, having three or more pregnancies, and having a live male child increase the odds ratio of unwanted pregnancies (15-17, 38). Despite the mentioned advantages, the logistic regression model is intensely affected by high correlations between the independent variables and the presence of nonlinear relationships (17).

In a study by Mohammadpourasl et al. (2004) on 1,576 women in the city of Tabriz, Iran, the prevalence of unwanted pregnancies was determined to be 26.7%. Using logistic regression, the factors affecting unwanted pregnancies were reported to be the high mother’s age and the number of live children3. Of course, in the present study, the mother’s age was not a significant variable in the logistic regression model, which could be due to the fact that only multiparous mothers were included in this study. This, in turn, limited the age range of the mothers. Still, the results obtained by the decision tree indicated the importance of the mother’s age in predicting unwanted pregnancies, especially among mothers with three or more pregnancies and a live male child.

In a study by Pourtaheri et al. (2007) on the women of Shahrood, Iran, the prevalence of unwanted pregnancies was 31%. Using logistic regression, the higher number of living children was offered as the risk factor in unwanted pregnancies, while in the present study, the number of live male children was one of the independent variables4. Moreover, a study by Amani et al. on 328 women in Ardebil, Iran, reported the prevalence of unwanted pregnancies as 60.7%. Using logistic regression, unwanted pregnancy showed a significant relationship with the parity, but not with the high mother's age, which matches the results obtained by the logistic regression model in the present study. However, in the present study, a high mother’s age was significant in the decision tree model. This issue could be attributed to the different nature of the decision tree in this study. That is, the impact of the age was only confirmed in mothers with three or more pregnancies and a live male live child, but not in all mothers (11).

A study by Vakili et al. (2011) on women living in Yazd showed that the prevalence of unwanted pregnancies was 24.5%, and based on the logistic regression model, the number of live children had a greater impact on unwanted pregnancies (26). The present study obtained similar results for the number of live male children based on the logistic regression and the decision tree models. Goto et al. (2002) conducted a study on Japanese women, and suggested that 46.2% of them had unwanted pregnancies. Logistic regression showed that higher parity and more live children had significant relationships with unwanted pregnancies (6). This result corresponds with the findings of the present study, although the prevalence of unwanted pregnancies was higher among Japanese women.

In a study by Calvert et al. (2013) on Tanzanian women using logistic regression, the relationship between a high mother’s age and unwanted pregnancies was found to be significant, which does not match the results obtained from the logistic regression in the present study (8). The reason for this difference could be the different combination of independent variables in the present study, as well as the inclusion of only multiparous mothers in this study. However, the results of the above-mentioned study match the results obtained using the decision tree in the present study, which showed a relationship between a high mother’s age and unwanted pregnancies.

Sadat-Hashemi et al. conducted a study in 2003 on women living in Tehran, using logistic regression, probit, linear discriminant analysis, and artificial neural networks, and the impact of factors such as the mother’s age and the number of live female/male children on unwanted pregnancies was determined. In the end, it was reported that the artificial neural network, especially the three-layer perceptron, was the best model, and discriminant analysis was the worst (25). In the present study, the predictive power of the artificial neural network was higher than the other methods, and that of discriminant analysis was the lowest.

In another study by Sadat-Hashemi et al. (2005) on women living in Tehran, the prevalence of unwanted pregnancies was determined to be 31.1%. Using artificial neural networks and multinomial logistic regression, and taking into account the mother’s age, number of live children, educational level of the couples, initial awareness of family planning methods, and the type of the contraceptive used before pregnancy, the predictive power of the two methods was compared. The artificial neural network was recommended as the better method. Moreover, regardless of the different nature of the variables (different categorization type), the superiority of the artificial neural network matches the findings of the present study (28).

Khalaj-Abadi-Farahani et al. (1996) conducted a study on 4,141 women living in Tehran, and the prevalence of unwanted pregnancies was reported as 31.1%. Using multinomial logistic regression, the variables of the high mother’s age and number of live male children were found to affect the unwanted pregnancies (9). The results of this study match the results obtained by the decision tree in the present study, but differ from the results by logistic regression. This apparent contradiction could be due to the different categorization of the dependent variable in the two studies.

In a study by Faghihzadeh et al. (2003) on women living in Tehran, the prevalence of unwanted pregnancies was 38.2%, and using log-linear models, significant relationships were reported between a high mother's age and low pregnancy spacing with unwanted pregnancies32. This finding corresponds with the results of the logistic regression model used in the present study about the impact of low pregnancy spacing, but does not correspond with the high mother’s age, which can be explained by the differences in the selected samples in the two studies.

The prevalence of unwanted pregnancies was reported to be 48% in a study by Finer and Zolna (2011) on women living in the USA. Due to the large sample size, no tests or models were implemented on the data. In this study, the unwanted pregnancies were higher among women younger than 25 years old and those having a higher parity (7). The relationship between a lower mother’s age and the higher prevalence of unwanted pregnancies contradicts the results of the present study, which can be attributed to cultural differences between American mothers and the mothers in this study.

Given the high prevalence of unwanted pregnancies among multiparous mother in Khorramabad, Iran, it is necessary to conduct further studies, revise family planning programs, and educate women who are at risk. It is recommended that the educational content be focused on selecting the ideal number of children and their genders based on the views of the couples, selecting adequate and proper pregnancy spacing, and the appropriate pregnancy age for the mothers. Given the role of different statistical models in determining the risk factors of unwanted pregnancies, and the difference between multiparous and nulliparous mothers, it is suggested that a separate study comparatively investigate the roles of the various risk factors among the women of the two mentioned groups.

Statistically speaking, it is suggested that the various classification techniques be compared in a simulation study based on criteria such as the type and number of independent variables, their distribution, samples size, and the type of architecture of the artificial neural network (including the number of layers, nodes, and type of activity function).

Footnote

References

  • 1.

    Moosazadeh M, Nekoei‐moghadam M, Emrani Z, Amiresmaili M. Prevalence of unwanted pregnancy in Iran: a systematic review and meta‐analysis. Int J Health Plann Manage. 2014; 29(3) : 277 -90

  • 2.

    Najafi F, Iran-far SH, Rezayi M, Iran-far KH. Systematic Review and Meta Analysis of Prevalence of Unwanted Pregnancy in Iran. J Kermanshah Unive Med Sci. 2011; 16(57) : 280 -7

  • 3.

    Mohammadpoorasl A, Rostami F, Ivanbagha R, Torabi S. Prevalence of unwanted pregnancy and multivariate analysis of its correlates in Tabriz city, 2004. Med Sci J Islamic Azad Unive-Tehran Med Branch. 2005; 15(4) : 201 -6

  • 4.

    Pourheydari M, Souzani A, Shamaiian N. Prevalence of unwanted pregnancies and their correlates in pregnant woman in Shahrood, Iran. Payesh. 2007; 6(1) : 63 -70

  • 5.

    Khousheh MG, Ebrahim TG, Hatami Z, Safaari M. Prevalence of unwanted pregnancy and its related factors in women referring to health centers in south of Tehran. 2008;

  • 6.

    Goto A, Yasumura S, Reich MR, Fukao A. Factors associated with unintended pregnancy in Yamagata, Japan. Soc Sci Med. 2002; 54(7) : 1065 -79 [PubMed]

  • 7.

    Finer LB, Zolna MR. Unintended pregnancy in the United States: incidence and disparities, 2006. Contraception. 2011; 84(5) : 478 -85 [DOI][PubMed]

  • 8.

    Calvert C, Baisley K, Doyle AM, Maganja K, Changalucha J, Watson-Jones D, et al. Risk factors for unplanned pregnancy among young women in Tanzania. J Fam Plann Reprod Health Care. 2013; 39(4) : 2 [DOI][PubMed]

  • 9.

    Khalajabadi FF, Sadat HS. Factors influencing unwanted pregnancies in Tehran. 2002;

  • 10.

    Kazemi SA, Kousha A, Taddayon P, Mousavi-nasab N. The Causes of Unwanted Pregnancy in 500 Pregnant Women Which Referring to Zanjan Hospitals. Journal of Zanjan University of Medical Sciences. 2000; 37 : 39 -45

  • 11.

    Amani F, Bashiri J, Moghaddam NN, Tabarraie Y. Application of logistic regression model in surveying effective causes of unwanted pregnancy. Qom Unive Med Sci J. 2010; 4(1) : 32 -6

  • 12.

    Shahbazi A, Ghorbani R, Akbarifar M. A survey on the prevalence of unwanted pregnancy and some related factors in pregnant women who referred to the medical laboratories in Semnan. Koomesh. 2006; 7(3) : 133 -8

  • 13.

    Mansoory A, Hossainy S, Dadgar S. Unexpected pregnancy and relative factors in pregnant women referring to Mashhad maternity wards in 2004. J Birjand Unive Med Sci. 2009; 16(1) : 65 -70

  • 14.

    Bayat H, Asefzadeh S. Prevalence of un-intended pregnancies and its relationship with low birth weight. J Birjand Unive Med Sci. 2004; 11(1) : 28 -31

  • 15.

    Dreiseitl S, Ohno-Machado L. Logistic regression and artificial neural network classification models: a methodology review. J Biomed Inform. 2002; 35(5) : 352 -9

  • 16.

    Harper PR. A review and comparison of classification algorithms for medical decision making. Health Policy. 2005; 71(3) : 315 -31 [DOI][PubMed]

  • 17.

    Melody YA. Comparative Assessment of Classification Methods. . Decis Support Syst. 2003; 35 : 441 -54

  • 18.

    Yassaee F. Prevalence of Unwanted Pregnancy for Women Which Referring to Mahdieh Hospital. J Med Res, Shahid Beheshti Unive Med Sci. 2001; 26(2) : 133 -6

  • 19.

    Abazari F, Arab M, Abbasszadeh A. Relationship of unwanted pregnancy and fertility behavior in pregnant women who visited maternity wards of Kerman hospitals. J Reprod Infertil. 2003; 4(1)

  • 20.

    Robabi H, Sarani H, Azarkish F, Dastfan Z, Dashipoor A. The survey of factors associated with unwanted pregnancy among women referring to health care centers of Iranshahr in 2007. Iranian J Obstetrics, Gynecol Infertil. 2011; 714(4) : 32 -40

  • 21.

    Rakhshani F, Moghadam AA, Tehrani H. Prevalence of unwanted pregnancy and associated factors in zahedan-1999. J Res Med Sci. 2003; 8(3)

  • 22.

    Khalili M, Shoohani B, Soltani O, Pour NA. A study on factors effecting unwanted pregnancy among the women referring to ilam health centers, 2003. J Illam Unive Med Sci. 2004; 12(42-43) : 18 -24

  • 23.

    Yazdani F. Comparison of the some factors in women with unwanted and planned pregnancy. Family Health. 2012; 1(2) : 19 -26

  • 24.

    Najafian M, Karami KB, Cheraghi M, Mohammad Jafari R. Prevalence of and some factors relating with unwanted pregnancy, in ahwaz city, iran, 2010. ISRN Obstet Gynecol. 2011; 2011 : 523430 [DOI][PubMed]

  • 25.

    Sadat HS, Kazemnejad A, Kavehei B. Use of various artificial neural networks for prediction of unwanted pregnancy and theircomparison using traditional statistical methods. J Shaheed Sadoughi Univ Med Sci Health Services. 2003; 11(1) : 10 -5

  • 26.

    Vakili M, Shahbazi H, Dehghan MH. The prevalence of unintended pregnancies and its related demographic factors in hospitals of Yazd city-2008. Toloo-e-Behdasht. 2011; 9(4) : 23 -36

  • 27.

    Norouzi A, Khoram Roodi R, Sharifi S, Tahmasebi R. Prevalence of unwanted pregnancy and its related factors in the women covered by health centers in Bushehr province in 2003. ISMJ. 2005; 8(1) : 83 -9

  • 28.

    Sadat-Hashemi SM, Kazemnejad A, Lucas C, Badie K. Predicting the type of pregnancy using artificial neural networks and multinomial logistic regression: a comparison study. Neural Comput Appl. 2005; 14(3) : 198 -202

  • 29.

    Nouhjah S, Niakan Kalhori SH. Artificial Neural Networks Application to Predict Type of Pregnancy in Women Equal or Greater Than 35 Years of Age. Int J Comput Eng Inf Technol. 2014; 3(1) : 177 -82

  • 30.

    Faghihzadeh S, Babaee Rochee G, Lmyian M, Mansourian F, Rezasoltani P. Factors associated with unwanted pregnancy. J Sex Marital Ther. 2003; 29(2) : 157 -64 [PubMed]

  • 31.

    Kitsantas P, Hollander M, Li L. Using classification trees to assess low birth weight outcomes. Artif Intell Med. 2006; 38(3) : 275 -89 [DOI][PubMed]

  • 32.

    Edrisi M, Gharipour M, Faroughi A, Javeri F, Shahgholi B, Gharipour A, et al. Decision support in prediction of metabolic syndrome with data mining methods. J Jahrom Unive Med Sci. 2011; 9(2) : 48 -58

  • 33.

    Toloiee-Ashlaghi A, Pourebrahimi A, Ebrahimi M, Ghasem-ahmad L. Using Data Mining Techniques for Prediction Breast Cancer Recurrence. . Iran J Breast Dis. 2013; 5(4) : 23 -34

  • 34.

    Sepehri MM, Rahnama P, Shadpour P, Teimourpour B. A data mining based model for selecting type of treatment. Tehran Unive Med J TUMS Publications. 2009; 67(6) : 421 -7

  • 35.

    Keshtkar A, Majd-Zadeh SR, Mohammad K, Ramezan-Zadeh F, Borna S, Azemi-Khah A, et al. Determination of effectvie factors on preeclampsia severity the application of classification and regression trees. J Gorgan Unive Med Sci. 2006; 8(2) : 47 -54

  • 36.

    Saki Malehi A, Hajizadeh E, Fatemi R. Evaluation of prognostic variables for classifying the survival in colorectal patients using the decision tree. Iranian J Epidemiol. 2012; 8(2) : 13 -9

  • 37.

    Hosseini SM, Tazhibi M, Amini M, Zaree A, Hashemi HJ. Using Classification Tree for prediction of Diabetic Retinopathy on Type II Diabetes. J Isfahan Med School. 2010; 28(104)

  • 38.

    Tu JV. Advantages and disadvantages of using artificial neural networks versus logistic regression for predicting medical outcomes. J Clin Epidemiol. 1996; 49(11) : 1225 -31 [PubMed]

  • COMMENTS

    LEAVE A COMMENT HERE: