Outcomes of Nfp Home Visitation 1

OUTCOMES OF NFP HOME VISITATION 1

On-line Supplement

This supplement contains four sections. The first provides demographic data on clients NFP served through 2013. The second describes how we estimated NFP’s impact on preterm second and subsequent births. The third reviews mixed, negative, or as yet insufficient evidence onseven additional or longer-term NFPoutcomes. The fourth describes the methods underlying our estimate of the Medicaid cost savings associated with NFP outcome improvements and its standard error.

Profile of NFP Participants

Online Table 1 summarizes selected data from the NFP national data system. Of note, the US trials assessed NFP effectiveness in white, black, and Hispanic populations. Those populations account for 92.8% of all clients served.

Pooled Data Tables

Table3in the main text providesdata by trial and pooled best estimates for most evidence-based binary outcomes of NFP. The table uses numbers computed by multiplying case counts times published percentages, often resulting in fractional case counts. We sometimes would have obtained integer estimates had more decimal points been published, but sometimes the non-integer numbers result from statistical adjustments to equalize control and treatment groups characteristics at program entry.

The remainder of this section adds details about a few entries in Tables 3 and 4.

In the preterm birth data in Table 3 and the low birthweight data in Online Table 2, only the first-born of multi-parous births generally is counted. Olds et al. (1986) excluded 24 births based on decisions of a pediatrician co-author who was blinded to group number. Regression-adjusted preterm birth counts for Denver, Elmira, and Memphis came from personal communication with MichaelKnudtson and David Olds. Dutch preterm birth case counts came from personal communication with Jamila Mejdoubi; another 34 controls and 18 treatments were ‘unable to judge if preterm’. Louisiana counts were inferred from the sample sizes and the reported percentage reduction and its significance.NFP National Service Office tracking data showed that nationally 9.3% of births were less than 37 weeks gestation for mothers enrolled in NFP at the time of delivery in 2005-2007. (These data include 89% of all live births to NFP enrollees who resided in NFP catchment areas at the time of birth.) By comparison, the age-matched national average preterm rate shown in Table 3 was 13.3%, a 30% difference (National Center for Health Statistics, n.d.). For the 24% of NFP mothers under age 18, the difference was 39% (9.2% vs. 15.0%). For mothers ages 18 and over, it was 19% (9.3% vs. 12.7%). NFP preterm rates by maternal age group were lower than national averages for black, Hispanic, non-Hispanic white, Asian (using more years of data to get adequate cases), and Native American mothers. The 30% prematurity rate difference is conservative. It would grow if the non-NFP group could be restricted to mothers on Medicaid or to first births (Institute of Medicine (US) Committee on Understanding Premature Birth and Assuring Healthy Outcomes, 2007).

The Oklahoma and Cincinnati infant mortality rate reductions in Table 3 are suspect because the comparison groups are not equivalent. Program targeting means NFP mothers are at higher risk than most comparison mothers, but they probably are at lower risk than mothers who decline offered NFP services.

In analyzing intimate partner violence (IPV), we adjusted for recall bias using multipliers of 1.073 for one year recall versus 6 month recall (Bushery, 1981), 1.30 for 3 years versus 1 year (Borowitz, 2010), and 1.50 for 6 years versus 6 months (Bradburn, Rips, & Shevell, 1987). We used Denver surveys at ages 4, 6, and 9 that gathered IPV in the past 6 months and the past 3 years to estimate age-specific multipliers to estimate 6-month IPV rates from 1 year, 3 year, and 6 year rates. The multipliers accounted for the likelihood that the same person was victimized over multiple 6-month periods during the longer exposure period. Denver estimates before age 4 had never been analyzed until we inquired about them.

Gains in language development shown in Table 4 occurred almost exclusively among children of mothers who were more psychologically vulnerable. At least in Memphis, those gains improved achievement test scores in reading and math during grades 1-6. (Olds et al., 2010)

Computation of Preterm Second and Subsequent Births

Our computations had three stages. First we computed preterm birth rates by marital status, then used data on marital status of NFP recipients to estimate how many of the closely spaced subsequent births avoided would have been preterm absent close spacing. Of all births in 2009, 12.2% were preterm and 41.0% were to unmarried mothers (Martin et al., 2011). Being unmarried raises odds of preterm birth by 1.46 (Shah, Zao, & Ali, 2011). So preterm birth rates are 10.3% (.122/(1.46*.41+.59)) for married women and 15.0% (1.46 * 10.3%) for unmarried women. Of 177,517 clients enrolled in NFP nationally through 2013, 15.9% were married at intake (NFP National Service Office, 2014). That suggests a 14.3% preterm birth rate for NFP clients absent NFP (.159 * 10.3% + (1 - .159) * 15.0%). The 14.3% rate applies to the 0.094 birth reduction (.0735/.782), meaning that before adjusting for close spacing or loss in replication, on average, 0.013 subsequent preterm births were prevented per woman enrolled in NFP.

Second, we adjusted for close spacing. Pregnancy spacing below 18 months raises risk of preterm birth by 1.92% for each month less than 18 months (Conde-Agudelo, Rosas-Bermúdez, & Kafury-Goeta, 2006). Applying the inter-pregnancy interval distribution for controls in Denver (Olds, 2010), the 0.094 births reduced had an average excess probability of being preterm of 18.9%. That adds .018 preterm births (.189 * .094) to the .0133, for a total of .031.

Third, we estimated the reduction in very closely spaced births beyond the second birth. In Memphis, pregnancies within 6 months of delivery declined 31% (10 percentage points) over 4.5 years (Kitzman et al., 2000). We netted out second births based on the assumption that the 43% share of prevented closely spaced second births within 24 months that were conceived within 7 months in Denver (Olds et al., 2010) applied. With that assumption, another .061 (.10 - .43 * .09) very closely spaced births were prevented over the next 2.5 years. These births have a 23% preterm rate (Conde-Agudelo et al., 2006), accounting for another .014 (.23 * .061) preterm births avoided. This brings the total reduction to .045. Multiplying that reduction times the 78.2% replication adjustment, the expected reduction is 0.035 or 35 per 1,000 NFP families.

Negative and Inconclusive Outcomes

Low Birth Weight: All six randomized trials examined NFP’s impact on birth weights less than 2500 g. Online Table 2 summarizes their findings. The direction of change is wrong-signed in three of six trials. Overall, although preterm births declined by 4.6%, this change was far from significant (p=.66).

Maternal Depression: The domestic violence reductions in Denver were accompanied by a 66% reduction in maternal depression at 9 years post-partum (Olds, 2010). No reduction was observed in Memphis at 9 years post-partum or in Elmira at 15 yearspost-partum (Izzo et al., 2005). Louisiana observed a 43% reduction in prenatal depression but no reduction 6 to 8 months post-partum (Nagle, 2002; Sonnier, 2007).

Miscarriage in Subsequent Pregnancies. As Online Table 2 shows, NFP mothers had fewer miscarriages than control mothers within 48 months of the first birth in all three trials. The pooled difference, however, was only significant at p=.14.

Intimate Partner Violence Beyond Age 4. IPV risk at child ages 6 and 9 was lower for NFP than control mothers in both Memphis and Denver (Online Table 2). The pooled differences were not statistically significant.

Grade Repetition. As Online Table 2 shows, children in NFP-visited families were more likely than controls to repeat a grade at school in both Memphis and Denver. The pooled increase was not statistically significant.

Maternal Criminal Offenses. In Elmira, maternal arrest and conviction rates for NFP mothers were 70% lower than for control mothers through 15 years post-partum (Olds et al., 1997). In Memphis, however, self-reported arrest and conviction rates were 36% higher for NFP mothers than for control mothers through 12 years post-partum (Olds et al., 2010).

Estimation of Reduction in Medicaid Spending

To model Medicaid spending impact, we multiplied the incident counts in Table 2 times unit medical costs. Online Table 3 shows sources and dollar amounts for the medical costs. Costs per incident were multiplied times percentage reductions from Table 2 to estimate savings per family. Some savings were adjusted to account for the probability of Medicaid graduation. The estimates use a government perspective and are based on average cost pricing. We inflated all costs to 2010 dollars and discounted future costs to present value at a 3% discount rate. We used bootstrap simulation with 990,000 iterations to estimate standard errors for the cost estimates.

Supplemental References

Adams, E. K., & Melvin, C. L. (1998). Costs of maternal conditions attributable to smoking during pregnancy. American Journal of Preventive Medicine, 15, 212-219.

Borowitz, J. (2010). The role of recall bias in estimating the “lemons effect".

Bradburn, N., Rips, L., & Shevell, S. (1987). Answering autobiographical questions: The impact of memory and inference on surveys. Science, 236(4798), 157-161.

Conde-Agudelo, A., Rosas-Bermúdez, A., & Kafury-Goeta, A. C. (2006). Birth spacing and risk of adverse perinatal outcomes: A meta-analysis. Journal of the American Medical Association, 295(15), 1809-1823.

Fang, X., Brown, D. S., Florence, C. S., & Mercy, J. A. (2012). The economic burden of child maltreatment in the United States and implications for prevention. Child Abuse and Neglect: The International Journal, 36(2), 156-165.

Institute of Medicine (US) Committee on Understanding Premature Birth and Assuring Healthy Outcomes. (2007). Preterm Birth: Causes, consequences, and prevention. Washington, DC: Natonal Academies Press.

Izzo, C. V., Eckenrode, J., Smith, E. G., Henderson, C. R., Cole, R., Kitzman, H., & Olds, D. L. (2005). Reducing the impact of uncontrollable stressful life events through a program of nurse home visitation for new parents. Prevention Science, 6(4), 269-274.

Machlin, S. R., & Rohde, F. (2007). Health care expenses for uncomplicated pregnancies. Rockville, MD: Agency for Healthcare Research and Quality.

Miller, T. R., Finkelstein, E., Zaloshnja, E., & Hendrie, D. (2012). The cost of child and adolescent injuries and the savings from prevention. In K. Liller (Ed.), Injury Prevention for Children and Adolescents: Research, Practice, and Advocacy (2nd ed., pp. 21-81). Washington, DC: American Public Health Association.

Miller, T. R., & Hendrie, D. (2009). Substance abuse prevention dollars and cents: A cost-benefit analysis (DHHS Pub. No [SMA] 07-4298). Rockville, MD: Center for Substance Abuse Prevention, Substance Abuse and Mental Health Services Administration.

Olds, D. L. (2010). Final Report from U.S. Department of Justice. Denver, CO: University of Colorado.

Olds, D. L., Robinson, J., Pettitt, L. M., Luckey, D. W., Holmberg, J., Ng, R. K., Isacks, K., Sheff, K., & Henderson, C. R. (2005). Age 6 test of home visits by nurses vs. paraprofessionals. Grant # 5 R01 MH062485-03, final report. Denver, CO: University of Colorado.

Pourat, N., Martinez, A. E., Jones, J. M., Gregory, K. D., Korst, L., & Kominski, G. F. (2013). Costs of gestational hypertensive disorders in California: Hypertension, preeclampsia, and eclampsia. Los Angeles: UCLA Center for Health Policy Research.

Shah, P. S., Zao, J., & Ali, S. (2011). Maternal marital status and birth outcomes: A systematic review and meta-analyses. Maternal and Child Health Journal, 15(7), 1097-1109.

Zhou, F., Santoli, J., Messonnier, M. L., Yusuf, H. R., Shefer, A., Chu, S. Y., Rodewald, L., & Harpaz, R. (2005). Economic evaluation of the 7-vaccine routine childhood immunization schedule in the United States, 2001. Archives of Pediatric and Adolescent Medicine, 159(12), 1136-1144.

Online Table 1. Demographic characteristics of women enrolled in NFP, 1996-2013

Characteristic / Number / % a
Hispanic or Latina / 43,827 / 27.8%
Not Hispanic or Latina / 114,034 / 72.2%
Refused/No Response / 19,661
Black / 41,274 / 30.5%
White / 75,214 / 55.6%
Asian/Pacific Islander / 6,628 / 4.9%
Native American / 3,141 / 2.3%
Multiracial / 9,030 / 6.7%
Refused/No Response / 42,235
Primary Language: English / 91,429 / 85.6%
Spanish / 12,840 / 12.0%
Other / 2,577 / 2.4%
Age at Enrollment: <15 / 5,039 / 2.8%
15-17 / 45,755 / 25.9%
18-19 / 46,575 / 26.3%
20-24 / 55,118 / 31.2%
25-29 / 16.135 / 9.1%
>=30 / 8,304 / 4.7%
Married at Enrollment / 26,849 / 15.9%
Enrolled by the End of 16th Week / 72,910 / 44.7%
Enrolled Weeks 17-28 / 80,738 / 49.5%
Enrolled after Week 28 / 9,460 / 5.8%
Potential Pregnancy Completers / 164,857
Retained through Delivery / 139,015 / 84.3%

a Percentages exclude refused/no response.

OUTCOMES OF NFP HOME VISITATION 1

Online Table 2. Estimates by randomized trial and pooled estimates for four dichotomous NFPprogram outcomeswhere pooled data do not show significant differences between NFPand control families

Domain
Community, First Author (Year) / Control Cases / % Yes / Treatment Cases / % Yes / Relative Risk Decline / 95% CI
Low Birthweight
Denver, M. Knudtson, personal communication, 2013 / 229 / 11.4% / 214 / 6.7% / 41.2% / [-8.8%, 68.2%]
Elmira (Olds et al., 1986) + Nonwhite / 160 / 4.4% / 192 / 7.8% / -78.6% / [-327.2%, 25.4%]
Louisiana (Sonnier, 2007) / 88 / 10.2% / 151 / 7.9% / 22.3% / [-77.0%, 65.9%]
Memphis (Kitzman et al., 1997) / 633 / 14.0% / 426 / 15.4% / -9.8% / [-47.5%, 18.2%]
Netherlands (Mejdoubi et al., 2014) / 177 / 11.3% / 204 / 12.3% / -8.5% / [-88.4%, 37.6%]
Orange County (Nguyen et al., 2003) / 85 / 10.6% / 69 / 5.6% / 45.2% / [-70.2%, 82.4%]
Pooled (p=.50) / 1372 / 11.7% / 1256 / 12.7% / -7.1% / [-25.1%, 15.2%]
Miscarriage Within 48 Months
Elmira low-income, Olds et al.(1988) / 54 / 11.0% / 28 / 3.0% / 72.7% / [-155.8%, 97.1%]
Memphis, Kitzman et al. (2000) / 443 / 12.0% / 203 / 10.0% / 16.7% / [-35.2%, 48.6%]
Denver, Olds et al. (2004) / 220 / 12.3% / 204 / 9.3% / 24.4% / [-31.7%, 56.6%]
Pooled (p=.14) / 717 / 12.0% / 435 / 9.2% / 23.3% / [-9.5%, 46.2%]
Intimate Partner Violence
a. Denver, age 6, M. Knudtson, personal communication, 2014 / 201 / 12.9% / 184 / 10.9% / 16.0% / [-45.3%, 51.4%]
b. Denver, age 9, same source as a / 190 / 7.9% / 170 / 7.6% / 3.1% / [-97.7%, 52.5%]
c. Memphis, ages 6-9, Olds et al. (2007) / 436 / 23.7% / 191 / 20.6% / 13.1% / [-20.4%, 37.2%]
d. Recall adjusted, 6 month / 436 / 11.9% / 191 / 10.4% / 13.1% / [-20.4%, 37.2%]
Age 6, pool a, d (p=.36) / 645 / 16.5% / 381 / 14.5% / 12.5% / [-18.1%, 35.1%]
Age 9, pool b, d (p=.43) / 626 / 10.7% / 361 / 9.1% / 15.1% / [-26.3%,42.9%]
Grade Repeated
Memphis, age 12, Kitzman et al. (2010) / 407 / 20.8% / 187 / 24.9% / -19.7% / [-63.7%, 12.5%]
Denver, age 9, Olds et al. (2014a) / 199 / 6.1% / 176 / 11.8% / -93.8% / [-281.2%, 1.8%]
Pooled (p=.32) / 606 / 16.0% / 363 / 18.6% / -16.1% / [-54.0%, 12.5%]

OUTCOMES OF NFP HOME VISITATION 1

Online Table 3. Unit medical costs by condition, sources for those costs, and Medicaid cost savings per family served (in 2010 dollars)

Category / Source / Cost/ Case / Savings/Family [95% CI]
Increased Child Graduation / Olds et al., 2010 / N/A / $2,046 [1,645, 2,447]
Fewer Closely Spaced Second Births on Medicaid / Henry J Kaiser Family Foundation, n.d.; Institute of Medicine, 2007; Machlin & Rohde, 2007 / $51,402 / $3,911 [2,761, 5,062]
Reduced Smoking While Pregnant / Adams & Melvin, 1998 / $224 / $11[0.4, 22]
Reduced Preeclampsia / Pourat et al., 2013, excluding preterm birth costs to avoid double-counting / $20,085 / $1,382[648, 2,116]
Reduced Prematurity / Institute of Medicine, 2007; Machlin & Rohde, 2007 / $35,388 / $616 [-104, 1,335]
Fewer Injuries / Miller, Finkelstein, Zaloshnja, & Hendrie, 2012 / $1,789 / $153 [94, 213]
Increased Immunization / Zhou et al., 2005 / $778 / $51 [-13, 114]
Reduced Child Maltreatment / Fang, Brown, Florence, & Mercy, 2012 / $12,190 / $2,178 [1,074, 3282]
Reduced Intimate Partner Violence / Miller & Hendrie, 2015 / $825 / $204 [16, 392]
Reduced Youth Substance Abuse / Miller & Hendrie, 2009 / $62 / $10 [4, 15]
Total / $12,540 [8,142, 16,939]