Modifiable behavioral factors in a biopsychosocial model predict inadequate and excessive gestational weight gain☆

Study population and sample 

The sample was recruited from the population of women who registered for obstetrical care over a 2-year period in a hospital and primary care clinic system serving a 10-county area of Upstate New York. Women were eligible for participation if they entered prenatal care before the third trimester, were age 18 years or older at the time of delivery, planned to deliver within the local hospital and keep the baby, and were healthy and mentally competent. In addition, only women who gave birth to live singleton infants are included in the research presented here. Seventy-two percent of the women recruited were eligible for participation. Twenty-five percent of women refused participation, leaving a final sample size of 622 women. Fifty-nine percent of the refusals (158) were tacit refusals, meaning that the women agreed to participate but did not return the informed consent by the time of the delivery. The population and sample subjects are primarily white (96%), rural, and socioeconomically diverse. Detailed characteristics of the sample subjects have been described elsewhere [7], [8]. All women gave informed consent.

Data collection 

Two methods of data collection were used. The women were mailed questionnaires in midpregnancy that collected data on the behavioral and psychosocial variables. Shortly after delivery, the women's obstetrical records were audited for data on the biological and sociodemographic characteristics. Body weight and height were measured following study protocols (empty bladder, in stocking feet and street clothing) by health care providers at antenatal visits. Weight was measured on Seca digital scales (Columbia, MD). Height was measured using a tape measure fixed to the wall and a moveable head board. Study procedures were approved by the Internal Review Boards of Bassett Healthcare (Cooperstown, NY) and Cornell University and were in accord with the Helsinki Declaration of 1975 as revised in 1983.

Determining gestational weight gain and BMI 

Gestational weight gain and early pregnancy BMI were calculated using the initial weight measured at the first prenatal visit for 547 (88%) women who entered the study during the first trimester of their pregnancies. Fifty percent of the first-trimester women were measured before the ninth week of gestation, and 77% before the eleventh week.

For the 75 subjects (12%) joining the study in the second trimester of their pregnancies, their initial weight was adjusted to the 9- to 11-week interval. The details and validity of the adjustment are described in a footnote.1

Conceptual framework and variables 

The underlying conceptual framework guiding the selection of variables to include in this study was the biopsychosocial model of George Engel [9], [10]. This model, based on Systems Theory, recognizes the hierarchical structure of all natural systems and the importance of behavioral, psychological, and social characteristics in understanding the health of an individual.

The outcome variable for this study is gestational weight gain measured in pounds. Gestational weight gain was used as a continuous variable and also as a categorical variable: gaining less than recommended, the recommended amount, or more than recommended according to the IOM gestational weight gain guidelines as shown in Table 1 ((1)).For obese women, the upper limit of the recommended range was set at 25 lb, the same value as that for high BMI women.

Table 1. Gestational weight gain characteristics of sample (N=622)

	Initial BMI and IOM classification	No. participants, n (%)	IOM recommended GWG (lb)	Mean GWG (SD)	Participantsa exceeding IOM guideline	Participantsa not reaching IOM guideline
					← n (%)→
	<19.8, low	9.0 (56)	28-40	30.5 (7.1)	11.1 (6)	38.9 (21)
	19.8-26.0, normal	49.4 (307)	25-35	32.7 (9.9)	36.9 (108)	19.1 (56)
	26.1-29.0, high	15.6 (97)	15-25	30.6 (11.7)	70.7 (65)	7.6 (7)
	>29.0, obese	26.1 (162)	At least 15b	23.2 (13.3)	41.0 (64)	26.3 (41)
	Total	100 (622)		29.7 (11.7)	40.8 (243)	21.0 (125)

BMI=Body mass index. IOM=Institute of Medicine. GWG=Gestational weight gain. SD=standard deviation. a27 preterm deliveries (<37 weeks) omitted. bRecommended range of 15 to 25 lb was used.

In this research, we examined three behaviors related to gestational weight gain: food intake, physical activity, and cigarette smoking. Given our interest in promoting more effective nutrition interventions in clinical practice, we measured variables in ways that were both valid and readily implemented in clinical practice. To assess physical activity, women were asked, ″How does the amount of physical activity you are getting now compare with your physical activity level before you got pregnant?” Their response options were: much less active, a little less active, about the same, a little more active, and much more active. The validity and reliability of this measure are described in a previous publication ((8)).

Several aspects of food intake behavior were included based on the literature ((11)). The proxy measure for energy intake was change in the amount of food eaten during pregnancy compared with the amount eaten before pregnancy. Women were asked, ″How has the amount of food you eat now changed compared with times when you were not pregnant?” The response categories were: a lot less food, a little less food, a little more food, and a lot more food. A similar measure, taken in the first year postpartum, was significantly related to change in kilocalorie intake from pregnancy to postpartum as measured by a food frequency questionnaire ((12)). Other aspects of food intake were assessed using additional behavioral questions, particularly the number of snacks consumed per day and the frequency of regular meals and breakfast. These items had been previously used by Ohlin and Rossner ((11)) and were found related to weight retention at 1 year postpartum. The number of glasses of milk and servings of fruits and vegetables consumed per day were measured as indicators of dietary quality. Alcohol intake in grams was assessed using the food frequency questionnaire.

The psychosocial variables used in this research were locus of control, self-efficacy, attitudes toward weight gain in pregnancy, feelings about motherhood, and career orientation. Measures for these variables have shown validity and reliability for the study of weight-related behaviors in pregnant women ((7)). In addition, an index of social support was created by averaging responses to two questions on the questionnaire that asked about the frequency of receiving practical help from friends and relatives and the number of people the woman could call on if she needed help. Possible scores ranged from 1 to 4.5. Sociodemographic and biological characteristics of the mothers, listed in Table 2, were obtained using standard items in the medical record and the mailed questionnaire.

Table 2. Sociodemographic and biological variables considered in the development of biopsychsocial model of gestational weight gain in pounds (N=622)

	Variable	Sample (%)	Meana GWG (lb)	Percentage above IOM GWG guideline	Percentage below IOM GWG guideline
	Age at delivery(y)
	<20	4.2	35.0	57.7	19.2
	20-<25	22.2	30.6	45.7	26.1
	25-<30	32.5	28.7	37.6	20.3
	30-<35	28.3	28.8	39.2	22.2
	35-<40	10.5	29.5	32.3	20.0
	≥40	2.4	36.2	60.0	0
	BMI in first trimester
	<19.8 (low)	9.0	30.5	10.7	39.3
	19.8-26.0 (normal)	49.4	32.7	36.5	20.5
	26.1-29.0 (high)	15.6	30.6	68.0	8.3
	>29.0 (obese)	26.1	23.2	42.6	25.3
	Parity
	0	41.3	32.1	48.3	18.3
	1	34.4	28.0	36.5	22.9
	2	15.6	29.5	34.0	19.6
	3	5.4	24.7	31.4	40.0
	4+	2.9	26.6	38.9	27.8
	Unknown	0.2
	Marital status
	Single	22.2	31.6	47.8	22.5
	Married	72.8	29.2	38.4	20.8
	Separated/divorced	5.0	28.6	41.9	29.0
	Education
	<High school	7.6	28.9	40.4	34.0
	High school graduate	31.2	30.1	41.2	19.6
	Some post–high school	26.1	28.6	43.2	24.7
	College graduate	21.7	29.4	37.0	21.5
	Post-college	13.5	31.8	40.5	13.1
	PCAP participant (≤185% PIR)
	Yes	42.8	30.6	47.4	19.9
	No	54.3	29.3	36.1	22.2
	Unknown	2.9
	Type of health insurance
	Private	26.4	30.5	39.0	19.5
	HMO	44.2	28.8	37.5	21.8
	Medicaid	28.5	30.5	47.5	22.6
	Unknown	1.0
	Advice on gestational weight gainb
	None	59.3	30.2	42.8	20.1
	Correct advice	31.5	28.5	37.2	26.0
	More than IOM guidelines	4.8	32.0	53.3	16.7
	Less than IOM guidelines	2.1	28.7	15.4	15.4
	Unknown	2.3

GWG=Gestational weight gain. IOM=Institute of Medicine. PCAP=Prenatal Care Assistance Program (New York's expanded Medicaid program for pregnant women). PIR=Poverty index ratio or the federal poverty line. aStandard error of mean (lb) ranged from 8-15. bObtained from the prenatal questionnaire; otherwise variables obtained from the audit of the medical record.

Control variables 

Three control variables were included in regression analyses predicting gestational weight gain to adjust for the timing of measurements during prenatal care. These are described in a footnote.2

Data analysis 

The first step in the analysis was to screen all potential variables individually for a relationship to gestational weight gain and to select those with a P value of <.15. For this step, each variable was entered into a regression model that included the three control variables. For some questionnaire items, response categories were combined because of the small number of respondents. The psychosocial scale scores were categorized into low, medium, and high scores based on the 33rd and 66th percentiles. Initial models for psychosocial scales also adjusted for BMI because this dramatically strengthened the association between gestational weight gain and the psychosocial characteristics. The variables selected at this step are shown in Tables 2-4 and are the only factors included in further regression analyses.

Table 3. Behavioral variables considered in development of biopsychosocial model of gestational weight gain in pounds (N=622)

	Variable	Sample (%)	Meana GWG (lb)	Percentage above IOM GWG guideline	Percentage below IOM GWG guideline
	Change in amount of food eaten in pregnancy
	A lot more	23.0	34.6	45.0	14.7
	A little more	61.4	29.6	40.6	21.2
	A little less	12.5	23.1	34.6	32.1
	A lot less	1.5	21.2	44.4	44.4
	Unknown	1.6
	Change in activity since pregnant
	Much less active	15.0	33.1	49.5	16.1
	A little less active	24.8	30.8	45.5	21.4
	Same	39.7	28.8	36.0	22.3
	A little more active	17.7	27.7	38.2	22.7
	Much more active	2.7	27.6	35.3	35.3
	Unknown	0.2
	Smoking status in pregnancy
	Never smoker	51.4	29.4	37.8	22.5
	Past smoker	28.3	30.7	45.5	16.5
	Current smoker
	<0.5 pack/day	3.7	29.9	52.2	21.7
	0.5-<1 pack/day	7.9	28.4	36.7	26.5
	1-<1.5 packs/day	4.3	31.4	40.7	18.5
	≥1.5 packs/day	2.1	19.8	30.8	53.9
	Unknown	2.3
	Number of snacks per day in pregnancy
	None	1.3	26.6	37.5	50.0
	<1	19.3	27.3	38.3	22.5
	1-2	56.4	29.8	41.6	19.9
	3+	22.8	31.7	40.9	22.5
	Unknown	0.2
	Frequency of regular meals in pregnancy
	Never	0.8	25.8	20.0	40.0
	1 day per week	1.1	24.5	57.1	42.9
	2-3 days per week	6.6	30.1	39.0	24.4
	4-6 days per week	26.1	28.5	36.4	27.2
	7 days per week	65.3	30.3	42.6	18.2
	Unknown	0.2
	Number of fruits and vegetables per day in pregnancy
	Less than 1	3.1	30.8	47.4	15.8
	1-2	38.4	30.4	46.0	19.7
	3-4	45.0	28.8	37.5	23.2
	5 or more	13.3	30.3	34.9	21.7
	Unknown	0.2

GWG=Gestational weight gain. IOM=Institute of Medicine. aStandard error of mean (lb) ranged from 10-16.

Table 4. Psychosocial variables considered in the development of the biopsychosocial model of gestational weight gain in pounds (N=622)

	Variable	Sample (%)	Meana GWG	Percent above IOM GWG guideline	Percent below IOM GWG guideline
	Feelings about motherhood
	Low (≤3.57)	29.8	30.7	43.8	21.1
	Medium	39.5	28.8	37.1	22.0
	High (>4.0)	30.8	29.9	41.9	21.5
	Career orientation
	Low (≤2.46)	31.2	28.4	37.8	25.9
	Medium	27.4	31.0	44.4	16.6
	High (>2.69)	41.4	30.0	39.8	21.5
	Social support
	Low (≤3.0)	28.6	30.8	48.9	17.1
	Medium	52.0	29.1	37.2	21.9
	High (>3.5)	19.5	29.5	37.5	25.8

GWG=gestational weight gain. IOM=Institute of Medicine. aStandard error of mean (lb) ranged from 11-12.

Second, a series of multiple regression analyses began with the behavioral variables. A backward elimination scheme was used to reduce the factors to the most parsimonious model by removing factors not significant at the P<.05 level. This process was repeated with the addition of the psychosocial characteristics and the sociodemographic factors, resulting in a combined or full model. Again, nonsignificant variables were removed. Finally, all two-way interactions between each variable and BMI were statistically assessed. All models included the control variables. Observations with any missing information for the variables in the model were excluded.

The third step in the analysis examined the odds of gaining above or below the IOM guidelines relative to gaining within the recommended range for all variables previously described. Because there were many factors to consider simultaneously, the analysis proceeded in three steps. The initial individual assessment of each factor was made using a multiple generalized logit analysis that simultaneously adjusted for the trimester of the woman at the time of the prenatal questionnaire and the three control variables. All models excluded preterm (<37 week) pregnancies (n=27). A variable was considered for the next phase of model building if it was associated with one of the two logit models (inadequate or excessive gain) at the 15% significance level. Next, separate multiple logistic regression models predicting inadequate or excessive gain were developed. A backward-elimination selection procedure to determine the most parsimonious model was applied to each category of variables, behavioral, psychosocial, and sociodemographic. The criteria for a factor staying in the model during this phase was set to P<.05. The final models within each category were then combined together and reduced using the same procedure, except that BMI was constrained to remain in the model with the other control variables because of its importance in defining the outcome. Finally, all two-way interaction terms were assessed.

All analyses were performed on a personal computer using SAS software (version 7.0, 2001, SAS Institute, Cary, NC).

Gestational weight gain and its bivariate correlates 

The mean gestational weight gain varied by prepregnancy BMI group, with obese women gaining the least and normal BMI women gaining the most weight (Table 1). Exceeding the upper limit of the IOM gestational weight gain guideline was twice as likely as failing to reach the lower limit in the sample as a whole. However, across the BMI groups, there was a 2- to nearly 10-fold difference in the proportion of women with excessive and insufficient gestational weight gain. The low prepregnancy BMI group was the only group for whom insufficient gestational weight gain was more common than excessive gain.

Tables 2 through 4 show mean gestational weight gain and the percent of the sample gaining above the maximum and below the minimum limit of the IOM range for each of the behavioral, psychosocial, and sociodemographic variables considered for inclusion in the biopsychosocial model of gestational weight gain. Overall, the majority of behavioral variables and sociodemographic and biological variables were retained because they were significantly related to gestational weight gain in the expected ways. Half of the psychosocial variables (feelings about motherhood, career orientation, and social support) were retained. In general, the psychosocial variables were more weakly related to the outcome than the behavioral and sociodemographic variables.

A few findings merit particular mention. In this sample, only very heavy smoking (≥1.5 packs per day) appeared related to low gestational weight gain and inadequate gestational weight gain (Table 3). Advice on gestational weight gain was also related to excessive gestational weight gain (Table 2).

Multivariable biopsychosocial models of gestational weight gain 

The biopsychosocial model guiding this research calls for examining all of the factors listed in Tables 2 through 4 simultaneously. The final regression model was significant (F=13.9; P<.0001) and explained 27% (adjusted R2) of the variance in gestational weight gain. The regression model results are as follows. Decreased physical activity in pregnancy was associated with significantly greater gestational weight gain (2.74 lb; P<.01) than maintaining or increasing physical activity. Consuming much more and less food during pregnancy than prior to pregnancy were associated with greater (3.67 lb; P<.001) and less (−3.16 lb; P<.05) gestational weight gain, respectively, compared with maintaining similar levels of food intake. Women who consumed three or more servings of fruits and vegetables per day gained significantly less weight (−1.81 lb; P<.05) than those who consumed fewer servings during pregnancy. Women who smoked more than one and a half packs of cigarettes per day gained significantly less weight (−11.59 lb; P<.0001) than those women who did not smoke in pregnancy.

Interestingly, the effect of low social support varied significantly by BMI group (P<.05). Low social support among low, normal, and obese BMI women was associated with significantly more weight gain (2.81 lb; P<.01) than that of their counterparts with average or high social support. However, high BMI women who had low social support gained significantly less weight (−3.11 lb; P=.15) relative to high BMI women with average or high social support.

Several sociodemographic and biological variables also remained in the regression model with the behavioral and psychosocial variables. Low BMI women and obese BMI women gained significantly less weight (−3.30; P<.05 and −7.87; P<.0001) than normal BMI women. Age >40 years (7.26 lb; P<.01), parity greater than or equal to 1 (−4.08 lb; P<.001), and HMO health insurance (−2.32 lb; P<.01) were each significantly related to gestational weight gain.

The second research question addresses whether these same variables that are related to gestational weight gain in the continuous analysis are also related to the risk of both gaining above and below the IOM gestational weight gain ranges. As shown in Table 5, BMI and food intake were significantly related to the risk of both excessive and inadequate gestational weight gain. tableWomen with low early pregnancy BMI were significantly less likely than normal BMI women to gain above the IOM range and also significantly more likely to gain below the IOM range. The opposite was true for the high BMI women. They were approximately five times more likely than normal BMI women to exceed the top of the range, and only about one eighth as likely to gain below the recommended range. Women who reported eating much more food in pregnancy were 2.35 times more likely than women who ate a little more to gain too much weight in pregnancy and only about one third as likely to gain too little.

Table 5. Odds ratios for gaining above the IOM gestational weight gain guidelines versus in range and for gaining below the IOM gestational weight gain guidelines versus in range (adjusted)a (N=595)b

	Characteristic	Reference category	Above IOM guidelines vs In Range (n=458)		Below IOM guidelines vs In Range (n=353)
			OR [95% CI]	P value	OR [95% CI]	P value
	Low BMI	Normal BMI	0.22 [0.1, 0.6]	.002	9.50 [3.2, 28.6]	<.0001
	High BMI	Normal BMI	4.97 [2.7, 9.3]	<.0001	0.13 [0.1, 0.3]	<.0001
	Obese BMI	Normal BMI	1.84 [1.1, 3.0]	.02	0.70 [0.4, 1.3]	.3
	Much more food intake	Little more	2.35 [1.2, 4.5]	.01	0.29 [0.1, 0.6]	.001
	A little or a lot less food intake	Little more			2.39 [1.2, 4.8]	.01
	Income <185% PIR	>185% PIR	2.59 [1.6, 4.2]	.0001
	Interaction
	Income <185% PIR and much more food intake	Not both factors	0.33 [0.1, 0.8]	.02
	Less physical activity	Same or more	1.68 [1.1, 2.6]	.02
	Smoking ≥1.5 packs/day	<1.5 packs			7.63 [1.9, 30.2]	.004

IOM=Institute of Medicine. BMI=Body mass index. PIR=Poverty index ratio or the federal poverty line. aModel is adjusted for the trimester that the prenatal questionnaire was completed, the weeks of gestation, the weeks from the first to the last weight measurement, and the weeks from the last measurement to delivery. bPreterm births (n=27) are excluded.

Consuming less food in pregnancy and smoking one and a half packs of cigarettes per day during pregnancy were significantly related to risk of inadequate, but not excessive, gestational weight gain. Family income of less than 185% of the federal poverty line, the interaction of this variable and much more food intake, and physical activity were significantly related to excessive but not inadequate gestational weight gain. Women with family incomes less than 185% of the federal poverty line were approximately 2.6 times more likely to exceed the top of the IOM gestational weight gain range than women with higher incomes. However, as indicated by the significant interaction between low income and much more food intake, income was not as important an influence on gestational weight gain among women who reported that they increased their food intake substantially during pregnancy (OR=0.33). Women who were less physically active were 1.7 times more likely to gain more than recommended than those who maintained or increased their physical activity.