A Major Public Health Issue: The High Incidence of Falls during Pregnancy

Kari Dunning; Grace LeMasters; Amit Bhattacharya


Matern Child Health J. 2010;14(5):720-725. 

In This Article


Study Population and Recruitment of Subjects

Women were considered eligible if they had delivered within 8 weeks at one of six hospitals located in the Greater Cincinnati area, were residents of the tristate area (Ohio, Kentucky or Indiana) and were at least 20 years old. Public record birth certificate data from December 1999 through July 2000 identified 6217 eligible women who were mailed a letter describing the study and asking them to complete a telephone or internet survey. Names and addresses were matched to telephone numbers for 2810 (45.2%) women. Telephone interviewing began five days after initial letters were sent and at least eight attempts ondifferent days and times were made to contact subjects. Women who completed the 15 min phone or internet survey received five dollars. Those not responding by phone were then sent another letter and a four page mail survey including a two dollar bill. After this mailing, in 2 week intervals, a reminder postcard was sent followed by a letter including another mail survey. This study was approved by the University of Cincinnati Institutional Review Board and informed consent was obtained from all participants. Informed consent procedures for phone were verbal and for mail were written. For internet participants, the first screen explained the study and to consent they inserted a confidential unique identifier that had been supplied in a letter.

Definitions of Falls

To determine if a fall had occurred during pregnancy, subjects were asked: ''During this last pregnancy, did you experience any loss of balance, resulting in a fall where some part of your body––other than your feet––touched the ground?''

For participants with multiple falls, the most severe fall was analyzed defined by requiring medical attention or resulting in injury or restricted activity. For participants with multiple falls without medical attention, injury, or restricted activity, the fall that occurred later in gestations was chosen.

Telephone, Internet and Mailed Surveys

A fall and injury survey was developed with questions selected from other instruments[15–17] and received outside review from recognized injury experts. All instruments were pilot tested and silent monitoring was performed on 20% of phone interviewing conducted by experienced trained interviewers. The phone and internet surveys were identical. In order to maximize participation, the mail survey was limited to four pages and, therefore, included a subset of questions from the phone and internet questionnaire.

All surveys (phone, internet and mail) asked about demographics and social factors (often referred to as predisposing factors) including maternal age, race, presence of a permanent partner, problem with balance or vision prior to pregnancy, diagnosis of diabetes (mellitus and gestational), and employment status during pregnancy. Additional predisposing factors collected only from phone and internet surveys included maternal height, weight gain during pregnancy, number of toddlers age three and under cared for during pregnancy, exercise patterns prior to pregnancy, number of previous live births, desire to become pregnant, and the baby's weight and gestational age at delivery.

For those women reporting a fall during pregnancy, all surveys (phone, internet and mail) asked questions to estimate severity based on medical attention, injury and restricted activity. Medical attention included: none, doctor visit, emergency room visit or hospital admission. Injuries due to the fall included none, bruise, cut, turned ankle, sprain/strain, broken bone or other. Restricted activity (in days) included: 0, 1, 2–5, 6–10, >10.

In order to address fall prevention, detailed information regarding falls was collected (often referred to as situational factors). All surveys (phone, internet and mail) asked about month of gestation at the time of the fall, location, and whether the fall was associated with a slippery floor, uneven or sloped ground, bathtub/shower, stairs or a cluttered or poorly lit area. Type of shoe worn at the time of the fall was ascertained including heel height and if the shoes were slick, worn, loose, or backless. Phone and internet surveys asked about additional situational factors including: obstructed view, carrying an object or child, hurrying, falling >3 feet or pushed/struck (by accident or on purpose). In addition, during the phone/internet survey, the subject was also asked if she had experienced illness (including hypoglycemia, dizzy, and extreme vomiting or diarrhea) the day prior to the fall.

Assessment of Reliability, Validity and Participation Bias

Approximately 4 weeks after completion of the initial survey, a test–retest reliability study was undertaken on a 10% random sample of participants. Respondents were asked questions related to the occurrence and location of falls during their pregnancy. Validation of injury needing medical attention was attempted by requesting authorization to review medical records regarding the fall. Finally, in order to assess non-participation bias, women who did not participate in the phone, internet or mail survey were sent a self addressed stamped double serrated postcard asking if they had fallen during their pregnancy. In addition, birth certificate data was used to compare age (as a continuous variable) and birth hospital (public or private) between participants and non-participants.

Missing Data

The percents of missing data for each independent variable were less than five percent and evenly distributed across methods, except for race, which was missing in 4.0%, 2.0%, and 6.7% of the telephone, internet, and mail datasets, respectively. Imputation of missing data was done prior to regression analyses by supplying the mean or mode value for continuous or dichotomous factors, respectively. Missing values for race were replaced by a predicted value that was most probable by using the SPLUS function TRANSCAN.[18]

Data Management and Analysis

Telephone and internet data were entered directly into computer files. Mail survey data were computer entered using 100% keystroke double entry with 10% comparison against hardcopy.

Responses were standardized and coded and a system of checks and compare programs were applied as quality control measures. Descriptive statistics examined the distribution of each variable. Univariate analyses using chi square, unadjusted odds ratios, and 95% confidence intervals determined variables marginally associated with a fall. Logistic regression was conducted using the phone and internet participants because they provided more information compared to the mail survey. Model building proceeded by first entering variables at least marginally associated (P<0.25) with the outcome; those significant at P<0.05 were retained for the final model. Assumption of linearity was assessed and evidence of confounding and effect modification were examined by association between dependent and independent variables through stratified and logistic regression analyses. All statistical analyses were performed using SAS (Cary, North Carolina), version 8.1 and SPLUS 2000.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.