Risk factors for low birthweight in the public-hospitals at Peshawar, NWFP-Pakistan

Across the world, neonatal mortality is 20 times more likely for low birthweight (LBW) babies compared to heavier babies (≤ 2.5 kg)[1]. It is also established as an important risk factor for neonatal morbidity [2, 3]. The cohort of LBW (birthweight <2.5 kg) babies is likely to reflect two effects, namely a short gestational age (preterm births) and small for gestational age (SGA). Small for gestational age usually results from intra-uterine growth restriction (IUGR) [4]. However, if the mother is small, it may be normal for her to have a small fetus. In the current study gestational age is included as an explanatory variable, therefore the study is focused on the identification of risk factors for the complementary effect which is SGA.

In a literature survey, de Onis et al (1998) found that IUGR babies are at increased risk of perinatal mortality and morbidity, i.e. sudden infant death syndrome, poor cognitive development and neurologic impairment, cardiovascular disease, high blood pressure, obstructive lung disease, diabetes, high cholesterol concentrations and renal damage in adulthood [4]. Such babies remain a burden on government expense in developed countries and a permanent problem for their families in developing countries.

The incidence of LBW (<2.5 kg irrespective of gestational age) is estimated to be 16% worldwide, 19% in the least developed and developing countries and 7% in the developed countries. The incidence of LBW is 31% in South Asia followed by Middle East and North Africa (15%), Sub-Saharan Africa (14%) and East Asia and Pacific 7% [5]. Of the total estimated IUGR babies (<2.5 kg and = >37 weeks), Asia accounts for 75%, and with 20% and 5% born in Africa and Latin America, respectively. The IUGR accounts for 11% of the total babies in developing countries ranging from 2% to 21%, that is 6 times higher compared to developed countries [4].

In South Asia the incidence of LBW is 36%, 30% in Bangladesh and India, and 19% in Pakistan [5]. In Pakistan the LBW rate varies from 5% to 23% in different parts [6‐11], whilst IUGR in a community-based study in Karachi was found to be 24.4% [12].

Peshawar is the capital of the NWFP Province of Pakistan, where the health care facilities are used by the people of Settled areas of NWFP, and Federally Administered Tribal Areas of Pakistan (FATA or Tribal areas) including Afghan refugees since 1979. The diversity in area-status (Tribal/Settled), ethnicity (Afghan refugees/Pakistani) and differential in overall geo-demographics factors suggest a need to investigate LBW in Peshawar. The current study derives an explanatory multivariate regression model for LBW, that includes gestational age as an independent covariate. The remaining selected covariates are thus interpreted as explanatory of incidence of LBW under the category of SGA. Therefore, this prospective public hospital-based study in Peshawar focuses on LBW to investigate associated explanatory factors beyond gestational age. These factors contribute towards the explanation of the observed births that are SGA (including preterm and non-preterm SGA births).

Data were collected in a cross-sectional prospective survey on maternal and paternal geo-demographic factors, maternal health and pregnancy history (MHPH) and neonatal outcome from all public hospitals (Hayatabad Medical Complex, Khyber Teaching Hospital, Lady Reading Hospital, and Government Maternity Hospital) in Peshawar during August to November 2003 through clinicians on duty.

The research reported in the paper was approved by the Research Ethics Committee of Liverpool John Moores University, with reference number 02141. Approval was granted prior to the commencement of data collection by questionnaire. Verbal consent was obtained from each mother recruited for this study.

The data comprise questionnaire responses collected from 1,039 single birth mothers from a total of 2286 mothers delivered during the study period irrespective of birth-status and gestational age. The volunteer clinicians for this study collected data as per their usual shifts i.e. day time one week and night time the next. This gave us 45.5% of the available data. As far as possible all consecutive births during the duty periods were recruited into this study, resulting in only 4.5% records returned entirely empty. This may have been through refusal by the patient to take part in the study or an inadvertent omission on the part of the duty clinician.

Women were interviewed by the clinicians in local languages at admission in the reception, with the exception of emergency admissions, when they were interviewed in the labour room. The factors were recorded on a pre-designed questionnaire validated by health professionals during a pilot study in the same hospitals.

Five factors were collected as continuous measures, which were later banded into categorical measures according to the previous literature (Table 1 and 2). These measures include maternal age [14‐16], gestational age [16, 17], height [12, 18] and the gap between this and the previous pregnancy [19]. However, the threshold for maternal pregnancy weight during analysis was chosen as 57 kg for this study, as this value had the highest significance in the univariate analysis for LBW. For the purpose of this paper consanguinity was included under geo-demographic factors (Table 1). Pregnancy registration was taken as proxy for pre-natal care.

The category of LBW was defined as less than 2.5 kg [13]. Gestational age was calculated from the first day of the last menstrual period reported by the mother and categorised such that any delivery from 24 and <37 weeks were termed as preterm birth [13]. Some mothers may have had estimates of gestational age derived by ultrasound measurement made at antenatal visits, although this would not be the case for un-registered pregnancies which account for more than half of the observations.

The reliability of the estimates of gestational age was further investigated using a descriptive table showing the incidence of observations grouped by birthweight and gestational weeks, shown in the form of a normogram in Figure 1. This also show boundaries for the lower 10^th percentile (<10^th percentiles) and upper 10^th percentiles (>90^th percentiles) of birthweight.

A comparison of the profile of birthweight against gestational age with those listed in [20, 21] shows a remarkable level of agreement on inspection of the two sets of data. This is interesting to note as one set of figures was derived in a technologically advanced city from an industrialised country, and the second comes from a single hospital study in Karachi, while our figures refer to four public hospitals in a border province of Pakistan. This consistency was taken to indicate that the estimates of gestational age in the prospectively acquired data set are sufficiently good to justify a description of the LBW cohort as shown in the Table, comprising low weight births expected for gestational age and SGA, the latter including a sub group of IUGR. The analysis shown in the next section takes gestational age as an independent variable, so seek to explain the residual variations due to SGA.

From the covariates originally collected, twenty were used for the initial analysis, to keep at least 5 events (LBW) per factor, as there are 101 occurrences of LBW in the data. This ratio of events per variable is on the threshold of acceptability for the size of the pool of covariates prior to initiating the variable selection process, as recommended in the literature [22].

There were two levels of data analysis, Crude odds of LBW babies in different categories were calculated to identify the significant factors at univariate level [23]. Multivariate logistic regression [24, 25] was applied to uncorrelated groups of factors (Pearson's, Spearman's correlation with thresholds of 0.6), fitting the models using the software package SPSS [26] to assess the independent effects of factors on SGA [12] based on a backward stepwise approach. Wald's test was used for the significance of the factors in multivariate models [27] and chi-square was used for univariate models.

The interpretation of main effects and interactions was aided by calculating a one-way table of predicted means for each significant main effect, and a two-way table of predicted means for each significant (two-way) interaction, using the statistical package GenStat [28]. These were constructed by first forming a table containing fitted values for every combination of the seven factors in the model, then averaging over the factors that are not in the required table of predictions. This averaging was weighted using estimated population weights, formed by multiplying together a one-way table of weights for each factor, containing the proportions of cases recorded in each of its levels.

The results are for data from 1039 birth episodes, prospectively sampled in the four public hospitals in Peshawar. The overall incidence of LBW (<2.5 kg) was found to be 9.9%. However, the crude odds of LBW in all geo-demographic factors (i.e. area of residence, water sources, ethnicity, consanguinity, maternal age, family income, maternal and paternal education) (Table 1), and seven other factors from MHPH (i.e. gestational age, hypertension, anaemia, maternal pregnant weight, total abortion/miscarriage, abortion/miscarriage before this pregnancy, and maternal pregnancy registration), were found to be statistically significant at univariate level (Table 2).

In the total LBW babies, 41.6% (42/101) were preterm and appropriate for gestational age (≥ 10^th percentile, AGA), 17.8 (18/101) were preterm and small for gestational age (<10^th percentile, SGA) and 40.6% (41/101) were full term and SGA. Among the total LBW, 58.4% (59/101) were SGA, and 40.6% (41/101) were appropriate for gestational age (Figure 1).

This study has been conducted in public hospitals that cover only 9% (urban = 18%, rural = 6%) of the total births, whereas the majority 91% of the deliveries take place at home (78%), or in private hospital/clinics etc (13%) [29]. There is no proper system that can record their history at home and due to the non availability of databases this study had to recruit patients prospectively from the four main public hospitals in Peshawar. This necessarily excludes births in private hospitals, clinics and at home. Due to limited voluntary participation of clinicians, it was not possible to collect information from all mothers admitted for delivery in the four hospitals from August to November.

In the present study abortion/miscarriage was used to include induced and spontaneous abortion, "due to the taboos and sensitivity associated with reporting an induced abortion"[30]. It was not possible to interview women at their homes in privacy to separate induced and spontaneous abortion. It should be noted that pregnancy registration was used as a proxy for prenatal care. We were also unable to collect data on energy intake that might be one of the important factors in Peshawar as reported by others in developing countries [31].

Considering the reliability of the study, this study found that some of the factors, i.e. consanguinity, low family income, maternal and paternal education and non-fresh water areas, diabetes, hypertension, anaemia and abortions are comparable with other reports for population or population-based studies. However, the high ratio of mothers from Tribal areas 24% in this study, compared to the total female Tribal population 15% [32] in this region (NWFP plus Tribal area) may over estimate, whereas, the low proportion of teenage mothers 7.5% compared to compared to 19.6% [19] may underestimate the incidence of LBW in this study.

The motivation behind this study is to collect data on low birthweight in NWFP, starting from the four public hospitals available for the general public in Peshawar. This study provides baseline information and a start to debate low birthweight from public hospitals in this region, which could help with possible intervention regarding maternal and newborn health in the future. It is an observational study which generated a sample surveying all but a very small proportion of consecutive births attended by the clinicians who agreed to take part. While this reduced the overall sample size, the duty patterns alternated between day and night and so the stratification of sampling by clinician does not indicate any factor likely to introduce bias into the results of the study.

The overall incidence of LBW in this study at Peshawar (10%) was half that of recent studies (19%–23%) in Lahore and Karachi [7, 9] and the overall national average [5]. The differential in the incidence of LBW might be due to ethnicity/racial differences in Lahore and Karachi, compared to Peshawar. However, the incidence of LBW in Budhni village near Peshawar was reported to be 5% [6]. The variation in the present study and the Budhni village study could be due to the differences in the population based and hospital-based study. In the hospital-based study, the ratio of mothers at risk is suspected to be more prevalent than in a village study, due to referral of high risk mothers from the Basic Health Units (BHUs) based in villages. Another reason for the high incidence of LBW babies in our study compared to Northrop-Clewes study [6] could be the inclusion of mothers from Tribal areas and Afghan refugees. These mothers from Tribal areas and Afghan refugees were found to be at higher risk compared to mothers from Settled areas and Pakistani mothers, respectively.

Among the total LBW, 59% were SGA, and 41% appropriate for gestation (Figure 1). The main factors associated with SGA (maternal age, consanguinity, nationality, anaemia, and abortion/miscarriages) adjusting for gestational age are discussed as follows.

The incidence of low birthweight reported in this study is less than one-half of the overall figure for Pakistan. However, there is significant variation among groups of mothers with specific combinations of geo-demographic factors, maternal health indicators and pregnancy history.

The effect of anaemia and previous abortion/miscarriage in Tribal areas compared to Settled areas, and the independent effect of Afghan refugee status on birth weight need further investigation to identify the root causes of adverse pregnancy outcomes in Tribal areas and Afghan refugees.

Health awareness programmes delivered by health workers in the villages, at educational institutions and through a concerted media campaign regarding the effect of consanguinity, teenage mothers, abortions and anaemia could be helpful in reducing the adverse pregnancy outcomes.

The findings of this study are specific to public hospitals in NWFP-Pakistan. However they show value in registering with the hospital during pregnancy and potential value in improving data collection methods, perhaps by electronic means, to support the design of targeted public health interventions and monitoring of their effectiveness. Further studies on LBW in private hospitals and clinics, together with studies of home deliveries, are also needed in order to extend the scope of the present work on LBW in this region.