Recidivism Among Jamaican Teenage Mothers: A historical cohort study 1995-1998
...d - identified as the major wage earner - at the time of the first live birth in 1994, and at the time of the survey in 1998. Our findings not only illustrate the link between poverty and recidivism, but also underscore the enormous importance of maternal support for the reduction/avoidance of repeat pregnancies. Only 44 percent (66/151) of respondents from households in which the mother was the main wage earner at the time of the first live birth in 1994, and 34 percent (32/93) in which the mother is currently the main wage earner, experienced one or more repeat pregnancies during the study period. The proportion of repeat pregnancies was much higher among respondents from households in which they and their ±baby fathers°and/or spouse were the main wage earners at first live birth (82%), and at the time of the survey (70%). METHODS Sample Participants in this historical cohort study were females between the ages of 15 and 20 who had a first live birth in 1994, as teenagers between the ages of 11 - 16, in the parishes of Kingston & St. Andrew, St. Catherine and Manchester. The total number of births that occurred and were registered between January and December 1994 in these parishes was 1453: Kingston (746) and St. Andrew (60), St. Catherine (379) and Manchester (268) (Registrar Generals Department, 1998). Sample size The total sample size was estimated to be 325, based on a Chi square a priori analysis, with 95 percent power to detect significant differences between cases (program participants) and controls (non-participants). In anticipation of a 50 percent loss to follow-up because of the difficulties of finding individuals who might have migrated during the study period, a random sample of (N=650) birth records was abstracted from the Registrar Generals Birth Registration files. Procedures Data Abstraction One in every three birth records that met the selection criteria was scanned into a Laptop computer [Compaq Presario 1610] using Easy Photo Smart Page Pro. All of the information contained in the birth record was manually transferred to a data file in SPSS version 7.5. The identification number on the birth record was used as the identification number for each questionnaire, and the names and addresses were used to find the individual and to invite their participation in the survey. This unique identification number enabled us to link the self reports from the interview with the data abstracted from the birth record. Records from the Womens Centre were also reviewed to identify those of this cohort (N=650) who had participated in the WCJF program: the mothers age at delivery and the infants date of birth were the unique identifiers used to match the two records. Survey Every effort was made to locate and make contact with all members of the study cohort (N=650); and to invite them to participate in the survey. An attempt was made to trace each individual at no fewer than three addresses: the first identified from data abstracted from the Registrar Generals Department; the second and third being forwarding addresses within the study Parishes. Members of the target population who could not be located after such careful tracking, and / or those who migrated to Parishes other than the three included in this study, or out of Jamaica, were considered to be ±lost to follow-up°. Parental and individual consent forms were utilized to secure permission for participants to be interviewed. From this sample of 650, we were able to locate 266 individuals: 241 at the first address obtained from the Registrar Generals birth record, and 25 at forwarding addresses. One individual refused to participate in the study and five did not meet all the criteria for inclusion in the sample. The response rate was therefore 98 percent [260/266], and the final sample represented 18 percent (260/1453) of all births that occurred and were registered to the target population in 1994. This sample size allowed us to detect significant differences between cases (program participants) and controls (non-participants) with approximately 90 percent power. Questionnaire A structured questionnaire was developed and used for data collection. The questionnaire was constructed to measure variables reflecting the types of change the WCJF program was designed to influence: avoidance of repeat pregnancies, completion of primary and / or secondary education, and contraceptive use. Items used for measuring demographic and pregnancy related variables were derived from the 1993 Contraceptive Prevalence Survey of the Jamaica National Family Planning Board. Items used for measuring WCJF program related variables were selected to allow for comparisons with previous evaluation studies conducted by Brown, 1995, and Chevannes, 1996. Items used for the measurement of sexual risk variables were selected from the San Bernardino Teen Pregnancy Project managed by Dr. Susanne Montgomery of Loma Linda University School of Public Health. Drafts of the questionnaire were reviewed by experienced social researchers in Jamaica to ensure culture and language appropriateness, and the revised instrument was pretested using a random sample of 30 adolescent mothers from the target population. After incorporation of all suggested revisions, the questionnaire was again pretested in January 1998 with another random sample of 30 adolescent mothers. Final revisions were made prior to administration of the instrument to the study cohort. Variables The dependent variables for this study were (i) one or more repeat pregnancy; and (ii) the level of educational attainment as an indicator of enhanced quality of life. Independent variables were selected either to reflect the theoretical approach and effects of the WCJF program or because they were previously used to measure program outcomes. Demographic, birth record and WCJF data abstracted from records were validated and updated in the questionnaire. The variables that were measured include the following: Demographic variables: ethnicity, educational level, occupation, income, perceived socioeconomic status, type of union/marital status, composition of current household, place of residence, social group and religious affiliation; Family/ Household characteristics: number and type of household in which respondent grew up, whether mother and father were teenage parents, union/marital status, educational attainment, and occupation of mother and father, number of siblings, whether any siblings are teenage parents, composition of household in which respondent grew up, perceived socioeconomic status, and religious affiliation of respondents childhood home; Sexual risk variables: age at first sexual encounter, number of sexual partners, number of biological fathers of offspring; sex education, and use of contraception; Pregnancy related variables: age at each pregnancy, outcome of each pregnancy, reasons for pregnancy, and reactions to each pregnancy; WCJF and other support program related variables: type and sources of support received during each pregnancy including WCJF, program expectations, perceived impact/benefit of program participation, and perceived disadvantages of program participation; Other theoretically important variables: age of first sexual partner, age differential between respondent and each ®baby father, age, educational level and occupation of ®baby father at each pregnancy. Data Analysis The data was double entered into SPSS/PC version 7.5 and cleaned: checks were made for incorrect responses, missing values, and outliers. Descriptive statistics were calculated for all variables: frequencies, percentages and modes were performed on all categorical variables, and means, medians and standard deviations on continuous variables. Bivariate analyses were performed for all independent variables by repeat pregnancy status (RP+ / RP-) and by WCJF program status (WCJF+/WCJF-). Pearsons chi square tests and bivariate correlation were calculated to determine which of these variables were related to, and which variables were significantly correlated with the outcomes of interest, i.e. repeat pregnancy and educational attainment. Similar assessments were made between WCJF program status and the independent variables. Only variables that showed significant (p<.05) correlation with the dependent variables were included in the regression models. Survey data for the study cohort [n=260] was used to calculate the cumulative incidence (CI) of a repeat pregnancy among WCJF program participants compared to non-participants. The relative risk (RR) was used to estimate the magnitude of the association between exposure to the WCJF program (WCJF + / WCJF -) and the likelihood of a repeat pregnancy. It is expected that exposure to the WCJF program, through which participants are returned to school and provided with contraception, will reduce the risk of repeat pregnancies. Hence, the Attributable Risk Percent (AR%) was calculated to assess the public health impact of exposure to the program, given that the program components - education and contraception - are two widely recognized and essential interventions for delaying pregnancy. In addition, the Population Attributable Risk (PAR) and Population Attributable Risk Percent (PAR%) were calculated to estimate the excess incidence and percentage of repeat pregnancies that would occur if the program did not exist. These population measures were calculated using crude risk data. Incidence rate stratification was used as a technique to control for confounding in the analysis by measuring the association within homogeneous categories of each confounding variable. For all multivariate analyses, program status was a standard covariant to determine the effect of program participation on the dependent variable. The backward stepwise conditional logistic regression procedure recommended by Kleinbaum (1988) was used. RESULTS Characteristics of Repeaters Approximately 81 percent (110/136) of repeaters had their first live birth between 15 (54/136) and 16 (56/136) years of age; of the remaining 19 percent, 14 percent had their first born at age 14 (19/136) and five percent (7/136) at age 13. Fifty two percent (71/136) of repeaters lived in the Kingston Metropolitan Area (i.e. Kingston & St. Andrew), while 30 percent (41/136) were from St Catherine and 18 percent (24/136) from Manchester. Thirty two percent (44/136)were attending secondary school at the time of their first pregnancy, 27 percent (37/136) an all age school and 26 percent (35/136) high school: 58 percent (79/136) wanted to continue their education after the birth of their first child. The cumulative incidence of repeat pregnancy was 52.3 percent. The incidence of repeat pregnancy by maternal age at first live birth is given in Figure 1.The annual incidence of repeat pregnancy was 10 percent in 1995; 43 percent in 1996; 27 percent in 1997; and 10 percent in 1998. Twenty four percent (32/136) of repeaters were WCJF program participants compared with 76 percent (104/136) of non participants (Figure 2). Risk predictors for repeat pregnancy Participation in the WCJF program (p=.000); wanting to continue ones education (p=.000), and taking some action to continue ones education (p=.008) after the first live birth in 1994; use of contraception after first live birth (p=.015), and at last intercourse (p=.019); all showed significant (p<.05) inverse correlation with experiencing one or more repeat pregnancy. Compared to all other types of relationships, common law unions with either the current partner (p=.004) or their first ±baby father° (p=.009) showed significant direct correlation with experiencing one or more repeat pregnancy. On the other hand, the absence of a sexual relationship with their first ±baby father° (p=.006) was inversely correlated with the outcome of interest. Perception of ones current socioeconomic status as very poor and/or poor showed small but significant (p=.008) direct correlation with experiencing one or more repeat pregnancy. Households in which the mother was the major wage earner at the time of the survey (p=.000), and was also the major wage earner at the time of the first live birth in 1994, showed significant (p=.001) inverse correlation with the outcome of interest, whereas current households in which the respondent or spouse was the main wage earner showed significant (p=.000) direct correlation with one or more repeat pregnancies. The WCJF Program impact on repeat pregnancy Among WCJF program participants the incidence of repeat pregnancy was 37 percent (32/87) compared with 60 percent (104/173) among non-participants. Program participation reduced the crude risk of one or more repeat pregnancy by 61 percent (.45-.83). Non participants were 1.6 times (1.585, 1.24-2.02) more likely to experience a repeat pregnancy compared with program participants. The attributable risk percent (AR%) i.e. the risk of one or more repeat pregnancies among the sample population attributable to WCJF program participation decreased by 63 percent (-.63). This means that WCJF program participation reduced the incidence of one or more repeat pregnancy by 63 percent in the study population, and must therefore be considered protective. The population attributable risk (PAR), i.e. the excess risk of repeat pregnancies among both WCJF program participants and non participants was -0.0781. Therefore, if the program were to be eliminated the incidence of repeat pregnancies could increase by 781 per 10,000 population. The population attributable risk percent (PAR%) for this three year period from 1995 to 1998 was 14.92 percent. For the population of teenagers in the parishes of Kingston & St. Andrew, St. Catherine and Manchester, 16 years and under at the time of their first live birth in 1994, WCJF program participation reduced the incidence of repeat pregnancies over the three year period [1995-1998] by approximately 15 percent. In the stratified analysis where significantly correlated covariates were controlled for sequentially, the protective effect of WCJF program participation was enhanced (>=61%) for respondents who wanted to continue their education; were in visiting unions with their current partner or the father of their first live birth; and were unemployed at the time of the survey. The effect was slightly decreased (<61%) but still significant and protective for respondents who had their first live birth at age 14; used contraception after first live birth and at last intercourse; grew up in one household; had their first live birth in single parent female headed households or extended families; perceived the socioeconomic status of both their current household and household at first live birth as poor; had an average monthly income of under J$10,000; whose mother was the main wage earner in the current household; and lived in an urban area. In the multivariate analysis (see Table 1) all variables previously found to be significantly associated with repeat pregnancy were used as covariates and were controlled for simultaneously. Four variables retained their significant (p<.05) independent effects on the occurrence of one or more repeat pregnancy in this target population. Participation in the WCJF program was significant at the p=.01 level in models 1-3, and at the p= .05 level in models 4-6. Respondents desire to continue education after the first live birth and use of birth control at last intercourse were entered in model 2. Whereas use of birth control at last intercourse became increasingly significant throughout the analysis from p=.05 at entry into the model and p=.01 thereafter, the desire to continue education after the first live birth was significant in all except the final model i.e. model 6. Households in which the mother was the main wage earner at the time of the first live birth was entered in model 3 and increased in significance from p=.05 to p=.001 in the final model. Current households in which the respondent and/or spouse is the main wage earner was also entered in model 3 and increased in significance from p=.05 to p=.001 in the final model. Compared with non participants, the risk of one or more repeat pregnancy was 45 percent (.45, .22-.91) lower for respondents who attended the WCJF program. Respondents who used contraception at last intercourse also had a 43 percent lower risk (.43, .21-.85) of one or more repeat pregnancy than their counterparts who did not. The risk was also significantly lower (.34, .18-.66) for respondents who had their first live birth in households where a mother was the main wage earner, and increased (2.86, 1.45-5.50) for those who, in addition to their baby father or spouse, are the main wage earners of their current household (see Table 2). Table 1: r coefficients for backward stepwise conditional regression models relating various combinations of risk predictors to repeat pregnancy during the follow-up period [1995-1998] VARIABLE Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Sample Size 260 233 197 197 197 197 Constant .4103** 1.9013** 2.1606*** 2.0628*** 1.9859*** 1.7489*** WCJF -.9519** -.7501** -.8802** -.8387* -.8115* -.8033* bc_last sex -.6952* -.8654** -.9064** -.8928** -.8543** ceduflb: educ after FLB 1.0381** -.7692* -.7642* -.7618* -.7483 flbwage1: mother vs. others -.9372** -1.0794** -1.0488** -1.0677*** chwage1: mother vs. others -.3526 _ - - chwage6: R/BF/S vs. Others 1.0102* 1.2040** 1.0165** 1.0509*...