Tobacco Documents Online

American Journal of Public Health March 1992, Vol. 82, No. 3 Established 1911 Editorials 341 Wanting Children C. A. Miller 343 Preterm Birth, Intrauterine Growth Retardation, and Perinatal Mortality J. L Kiely and M. Susser ~.~ H .-a 3 Public Health Policy Forum (.~ N 1 o cn u,rn 346 Nutrition: The Need to Define "Optimal" Intake as a Basis for Public Policy Decisions E. L Wynder, J. H. Weisbwger, and S. K Ng I cn > o Lncn aO Commentary I .-1 0 N ~ in a f- 351 Snuffmg Tobacco Out of Sport G. N. Connolly, C. T. Orleans, and A. Blum : ?~- ~ to_Juu Featuring Children's and Adolescents' Health i- ~• ~ r~Jlu • ~ crlu:7 354 Prevalence of Obesity among Children of Military Dependents at Two Major Medical Centers C M. Tiwary and A. H. Holguin ~ fu!ZZO # IQ~F- # QF-YCC9 # a~IUZ 3 NG')V) H N HriS 358 Body Fatness and Risk for Elevated Blood Pressure, Total Cholesterol, and Serum Lipoprotein Ratios in Children and Adolescents D. P. Williams, S. B. Goinv T. G. Lohrnan, D. W. Harsha, S. R Srrrtivasan, L S. Webber, and G. S. Berenson r~ yYUrUtn .-+a _tr<:a OQ~tl-43 364 Childhood Chronic Illness: Prevalence, Severity, and Impact P. W. Newacheck and W. R Taylor 372 Reductions in HIV Risk-Associated Sexual Behaviors among Black Male Adolescents: Effects of an AIDS Prevention Intervention J. B. Jerrunott III, L S. Jenunott, and G. T. Fong 378 Birth Weight and Perinatal Mortality: The Effect of Gestational Age A. J. Wikax and R Slqcsrven 422 Morbidity among Pediatric Motor Vehicle Crash Victims: The Effectiveness of Seat Belts J. S. Osberg and C. Di Scala - AI'HA 426 Pediatric Gonococcal Infection, Florida, 1984 to 1988 J.-C. A. Desenclos, D. Ganity, and J. Wioten i l Amci= Public Health Aseaiatim 429 n At anta Evaluation of an Intervention to Reduce Playground Hazards Child-Care Centers J. J. Sacks, M. D. Brantley, P. Holmgreen, and R W. Rochat Featuring 432 Air Quality during the Winter in Qudbec Day-Care Centers S. Daneault, M. Beausolefi; and K Messing Child»en's and 435 Infant Mortality, Per Capita Income, and Adult Illiteracy: An Ecological Approach R Tirssenus, J. Canela, J. Alvare7, J. Sentis, and L Salleras Adolescents' Health 438 Which Newborns in New York City Are at Risk for Special Education Placement? D. Goldber& M. McLau,ghlin, M. Grasss; A. Tytwr, and S. Blum V 441 The Lowest Birth-Weight Infants and the US Infant Mortality Rate: NCHS 1983 Linked Birth/Infant Death Data M. D. Overpecly H. J. Hoffman, and K Prager 445 An Epidemiological Comparison of Mexican-American and White Non-Hispanic 8th- and 12th-Grade Students' Substance Use E. L Chavez and R C Swaim i TIMN 346468 (continued page 331)

Estimation of the Break-Even Point for Smoking Cessation Programs in Pregnancy Backgmund Successful pro- gramc to help pregnant women quit smoking have been ckve~ and evaluated, but formaI smoldngcessa- tion programs are nota part ofcare at most prenatal sites. The cost of such programs may be an issue. Consid- ering the costs of adverse maternal and infant outcomes resulting from sniokrtp„ we estimated there would be an amount of money a prenatal program could invest in smoking ces- sation and still "break even" eco- noinkally. Aledrads- A model was devel- oped and published data, along with 1989 hospital charge data, were used to arrive at a break-even point for smoking cessation programs in preE nancY- Results. Using overall United States data, we arrived at a break- even cost of $32 per pregnant woman. When these data were var- ied to fit specific US populations, the break-even costs varied from $10 to $237, with the incidence of pretettn low birth weight having the most itn- pact on the cost. C.onclusiarx It may be advis- able to invest greater amounts of money in a prenatal smoking cxssa t4on. program f+ar some poptttat3cuffi. However, for every pnpulation dm is an amount that catn be ktteaftd while still breaking even. (lm.XPkob- lic Heaith. 1992;SZ:~383-39EI) Miriam Shipp, MD, MPH, Mary S. Cmughan-Minihane, PhD, Diana B. Petitti; MD, MPH, and A. Eugene Washington, MD, MSc Inhodudion Cigarette smoking has been linked with adverse pregnancy outcome since at least the 1940s1,2 and is the most important cause of low birth weight (LBW) in de- veloped countries.3 However, smoking cessation during the first trimester leads to infant outcomes similar to those for non- smokers,4 suggesting that smoking cessa- tion programs during early pregnancy can alter pregnancy outcomes. Although successful programs to help pregnant women quit smoking have been developed and evaluated,54 formal antismoking programs are not a part of most prenatal care. The reasons why smoking cessation programs have appar- ently not been implemented are not known with certainty, but the cost of such programs is probably a major issue. A recent analysis9 of the cost-effec- tiveness of smoking cessation programs in pregnancy, however, showed that over $3 was saved for every dollar spent on smok- ing cessation. Their cost-effectiveness was also demonstrated in a study that ex- amined such a program in a health main- tenance organization (HMO) popula- tion.1o These two studies looked at cost- effectiveness by assuming a fixed program cost and comparing this cost with that of the adverse medical outcomes. An alter- native approach to cost-effectiveness analysis is to examine the cost of adverse medical outcomes with an intervention and compare it with the cost without such an intervention. The difference in these two costs is the amount that the program could invest in the intervention and still "break even" economically. Our analysis takes such a perspec- tive. We calculated the aggregate costs of medical outcomes for a hypothetical pre- natal program that included a formal smoking cessation intervention and com- pared them with a hypothetical prenatal program that did not have such an inter- vention. Assuming that any program must decide how to allocate fixed resources for a group of pregnant women, who com- prise both smokers and nonsmokers, we therefore arrived at an estimation of cost per pregnant woman rather than of cost per pregnant smoker in our final result. Then, after arriving at this break-even point, we examined the extent to which this amount might vary from program to program, depending on the prevalence of smoking, the success of the cessation pro- gram, the incidence of adverse pregnancy outcomes in the population in question, and the cost of medical care. Methads Decision Analysis Models , We constructed two decision trees---one for infant outcome (Figure 1) and one maternal outcome (Figure 2}-to compare two strategies: (1) providing a The authors are all at the School of Medicine of the University of California at San Francisco- Miriam Shipp with the Institute for Health Pol- icy Studies, Mary S. Croughan-Minihane and Diana B. Petitti with the Department of Family and Community Medicine, and A. Eugene Washington with the Institute for Health Policy Studies and the Department of Obstetrics, Gy- necology, and Reproductive Sciences. Requests for reprints should be sent to Dr. Mary Croughan-Minihane, Department of Family and Community Medicine, University of Calffornia, Box 0900, Rm AC-9, San Fran- cisco, CA 94143-0900. This paper was submitted to the journal February 22, 1991, and accepted with revisions July 10, 1991. TIMN 346469 March 1994 VoL 82, No. 3 American Journal of Public Health 383 T TV ro :4Ek V"''Ts

Shipp et aL Program Mother smokes at first visit Quits smoking Does not quit Mother doesnY smoke at first visit Pretenn Term, LBW Neither Pretemi Term, L8W Neither Pretenn Term, LBW Neither Pregnant Women Usual Care Mother smokes at first visit Quits smoking Does not quit Preterm Term, L8W 0 Neither Preterm Term, LBW Neither Mother doesnR smoke at first visit Preterm Term, LBW 1 Neither FIGURE 1-Decisfon tree depicting infant outcome for pregnant women provided or not provided with a formal artwking cessation program. Circles are chance nodes and square ia a decision node. i Program Mother smokes at first visit Quits smoki Does not quit Mother doesn't smoke at first visit AWuptio. hemorrhaqe. previa Neither Pregnant Women Mothersmokea at first visit Quits smoki Do« not quit Abniplio. hemorrhaqe, previa ~~• ~ Neither AbrupGo, hemormaqe. pr.via Preeaartqsie NNtha ADtiptio. n.nnnh,O.. pr.via R/other doant smokw at first vidl e Preeclamollis Neilh« FIGURE 2--Declsion tree depicting mabrral outcort>• for pregnant women providad or not provided with a formal smoldng cessation proyram. Data and Assumptions Prevalence of Smoldng at First Pre- natal V'rsit (Node 1). Population-based data on smoking prevalence were ob- tained from the 1985 National Health In- terview Survey (published in 1988)." Among survey respondents, 31.8% of the women between the ages of 18 and 44 who had given birth to a child within the past 5 years smoked cigarettes during the 12 months preceding pregnancy. Because •21.2% of the women quit smoking upon learning they were pregnant, we estimated a baseline probability of 25% for women still smoking at the time of their first pre- natal visit (.318--(.318) (.212) _ .25). This prevalence of .25 is the figure used for all the baseline analyses. The greatest vari- ability in smoking prevalence was for women with different education levels. Therefore, probability estimates for the sensitivity analysis were based on a com- parison of women with fewer than 12 years of education (probability .39) and women with 16 or more years of education (probability .08). These estimates take into account the before-pregnancy smok- ing rates and the chances of women quit- ting upon leaming of pregnancy. Quit Rate after First Prenatal Visit in Pregnant Smokers Who Are Provided with a Formal Program (Node 2). Ran- domized clinical trials of smoking cessa- tion programs allow us to estimate quit rates attributable to formal smoking ces- sation programs. An optimal intervention program for pregnant women was imple- mented and evaluated by Sexton and Hebel,7 who achieved the highest quit rate among pregnant smokers published to date (43%). Subtracting the quit rate in the control group (20%) in that study from the quit rate in the intervention group, we es- timated a quit rate attributable to the pro- gram of 23%; this figure was used for the baseline analysis. In a study of a different smoking cessation program by Windsor et al.,6 the quit rate attributable to the pro- gram was 12% (14% intervention, 2% con- trol), which provided the lower probabil- ity estimate for our sensitivity analysis. We obtained the upper probability esti- mate of 29% by taking the difference in the smoking cessation rates between the in- tervention (43%) and control (14%) groups for men who participated in the Multiple Risk Factor Intervention Trial (MRFIT)12 study. Although the MRFIT study evalu- ated men rather than pregnant women, we formal smoking cessation program for program. Probabilities used in the deci- used it as the high estimate of the effec- pregnant women who smoke, or (2) not sion model were estimated based on pub- tiveness of a smoking cessation program providing a formal smoking cessation lished data (Table 1). because MRFTI's smoking intervention Abruptio, hemorrhage, previa Pre^.dartpsia Neitha Abruptio, hemorrhage, previa -~ ve,.,wV&a Neither 384 American Journal of Public Health TIMN 346470 March 1992, Vol. 82, No. 3

t i TABLE 't-4+talW69lks tb"al W illochtoctAn.lflrsk 8aanik0 Raq® af Prabar:liKtY ProbeAty Node Saxoe Prevaiertae o! atnoftiEtg atllrat pr+arwld visit .26 .08-39 1 P1oiS,1988' " tae kt pt~tteit~ Wnp[aetit"` w#q alft ptvkii9d .2i .12-29 2 Seodon attd Piebet,1984' with a ftltTCtllt prowaIrt Y1ideor et al.,19856 MRFff 1862 Spontaneous qtlit rate M't pregnerrtt srttokets" .02 .0t3-.08 3 VNindeor et al., 1885° MK.'14flttlx 6't al., 198713 Probebility af LSW (<2600 g).:3t 1erm (a37 wks) notwttokers .017 .015--.031 6. t~HS 1888„ t~tit smokittg aif~r 1 st vfsit .017 .015-031 4 PuBer and 8errano,198714 Otatsted at ai.,1985's o0titlntta Stltokinp 8t89r 9etvis& t)90 .t)53-.109 5 McNiOostt,1984te Probability af LBW (<2500 q) anct ptpkrm (<37 w(m) nonsmokers .035 .Q29-.086 8 NG#iS,1988„ quit stttokfnq after 1ot vbt .035 .t7129--.t>86 4 F'u%r and Setratto,1987''` Ktarner,1987° cm*w smoift after ist visit tM8 .041-.0H2 5 McIntosh,19641° ProbebiNty of notmAi birttt w®igttt (~2500 g) at term (2-37 wks) nonsmokers .948 .903-.958 6 NCHS 198811 quit arttokln9 after 1st vlsit .948 .904-.$56 4 Pull9r arxf Serrano, 1987,a Ounsted ®t 81.,1%'i'a tontlnue t;tnoking after 1st visk .891 .796-.906 5 Ksnrner,1967° ProbelMy+ o€ abnipf3o ptswtta, ple=ft ptnvi:3, or argoWmxn harnorrttage nonsmokers .009 .007-021 9 Meyer et sd.,1978x' quit wdklrtg after 1st visit .009 .037-.021 7 Reid et ai.,19r2s CuntlRtyhOrn et al.,19W contknae arttokirtp aRer 1st vEsit .013 .010-.029 8 AAcIn/oett,1984t6 Probabifty of preeclatnpsia of pr",ercy nonsmokers .0b8 .058-117 9 Marcoux et al., 19692' quit 8molft after 1et visit .058 .068-.117 7 ai., 19~'` Ctxriingitem e t ~ ,~ , ~ ~,vG - CioplBtud,la cottfxxje smokhg after 1st visit .024 .024-.049 8 Ptobebilit}r of none of the 8batne cotnpit:eGfona nonsmokers .933 .862-.936 9 Marcoux et al.,198927 quit smokttg after 1ed visit .933 .882-.906 7 Ctrtnktghatn et al., 19W Gopierud,18822° continue srrtokkV atter let vleit .963 .922-.968 8 Fiei~d et e1.,197225 M®yer et al., 197B?3 'wmnen wtw >xnoke at the fhat prenAfal visL was highly effective and may represent the maximum achievable for a formal cessa- tion program. There is no assurance, how- ever, that the same results could be achieved in pregnancy. Spontaneous QuitRateafterFkstPtre- nata! V'uu in Pregnartt Smokers (Node 3). Spontaneous quit rates were assumed to be the same as quit rates in women in the con- trol groups of randomized clinical trials. The baseline probability (.02) was derived from the 2% quit rate found among control group women participating in the study by Wind- sor et a1.6 The upper probability range (.06) was derived from a study by MacArthur et March 1992, Vol. 82, No. 3 al.,13 and the probability range of zero was chosen as an alternative, assuming that all 'Regarding the estimates for nodes 2 and 3, we used data from randomized trials of smoking ces- sation programs to determine the net effect at- tributable to the smoking cessation program, which was estimated as the difference between the quit rates in the intervention group and the control group. This was the estimate used for the program arm of the decision tree. In the nonpro- gram arm of the decision tree, we also used this data on quit rates from the control groups to es- timate the community spontaneous quit rate be- cause these studies are among tite few reliable sources of data on quit rates among pregnant wotnen. women who are going to quit smoking do so prior to their first prenatal visit.' Infant Outcomes Probability of LBW (< 2500 g) at Term (> 37 weeks) for Nonsrnokers and Quitters (Nodes 4 and 6). Only incidence rates and relative risks were available in the literature. Thus, we used the following formula to calculate the probability of LBW at term for nonsmokers and quit- ters: IL.aw PNS = (RRs) (PVs) + (PVNS) TIMN 346471 e American Journal of Public Health 385

Shipp et al. where PNS = the probability of the out- come among nonsmokers and quitters, ILBW = the incidence of LBW at term for the population, RRs = the relative risk of LBW at term for continuing smokers, Pvs = the prevalence of smoking for the population, and PvNS = 1-Pvs. The incidence rate of LBW at term (2.8%) was derived from US population- based data,l' and the relative risk of term LBW for smokers (3.5) was obtained from a study by Ounsted et al.15 The value used for the prevalence of smoking at the first prenatal visit was .25, and was derived as explained earlier. Probability of LBW (< 2500 g) and Pretenn (< 37 weeks) for Nonsmokers and Quitters (Nodes 4 and 6). The same formula was used to calculate the proba- bility of LBW and preterm for nonsmok- ers and quitters, substituting a baseline overall incidence of preterm LBW of 3.8% and a baseline relative risk of preterm LBW of 1.4.3.14 Probability of LBW (< 2500 g) at Term (? 37 weeks) and Pretenm (< 37 weeks) for Continuing Smokers (Node 5). The probability of each outcome among continuing smokers was derived from the probability of the outcome among non- smokers and quitters and the relative risk of the outcome for continuing smokers: PNon„ = the probability of normal birth weight at term, Ple,e1,,, = the probability of LBW and preterm, and Pr~,,,, = the probability of LBW at term. Sensitivity Analysis for Infant Out- come. Table 1 presents the range of prob- ability estimates used in the sensitivity analysis. The first sensitivity analysis var- ied the prevalence of smoking at the first prenatal visit, using the lower (.08) and upper (.39) probability range values to de- termine the individual effect of smoking prevalence on the break-even point. Ad- ditional analyses were then conducted us- ing the probability range estimates for each additional variable, varying only one estimate at a time. The analysis for the variable preterm LBW had values ranging from a low incidence of 3.2% to a high incidence of 7.3%, and the values for LBW at term were 2.4% and 5.0%, re- spectively.14 Because exact probabilities of infant outcomes could not be obtained from the literature and because the model appeared to be particularly sensitive to the relative risk of preterm LBW, a second sensitivity analysis examined the independent effects of varying the incidence rate for each out- come and relative risk estimate for pre- term LBW. A final sensitivity analysis ex- amined the independent effects of varying the incidence rate for each outcome and relative risk estimate for term LBW. obtained from a recent case-control study,27 which obtained a relative risk value similar to that obtained in older stud- ies.19 The reason for the lower relative risk of preeclampsia in smokers is uncertain, but it has been theorized that nicotine might inhibit the potent vasoconstrictor thromboxane, a substance found to be in- creased in studies in preeclamptic wom- en.28•29 For the probability of preeclampsia in •continuing smokers (node 8), we multi- plied the relative risk in smokers (.42) by the probability of preeclampsia in non- smokers and quitters (PNS). Probability of Births Not Compli- cated by Placenta Previa, Placenta Abruptio, Hemorrfzage, or Preeclampsra (Nodes 7, 8, 9). The probability of a preg- nancy not complicated by any of the above diagnoses was derived by subtrac- tion. Sensitivity Analysis for Maternal Outcomes. For the range of probabilities cited in Table 1, we varied the incidence rates of hemorrhage and preeclampsia through the ranges cited in the literature (.008 to .023 for hemorrhage and .05 to .10 for preeclampsia) while maintaining the relative risks for continuing smokers. Additional sensitivity analyses var- ied, one at a time, the prevalence of smok- ing at the first prenatal visit, the probabil- ity of quitting smoking with a smoking program in place, and the probability of quitting smoking without a program in place. PS = (Prrs) (RRs), where Ps = the probability of the outcome among continuing smokers, PNS = the probability of the out- come among nonsmokers and quitters, and RRs = the relative risk of LBW for continuing smokers. The range of relative risk estimates for continuing smokers was obtained from the Alameda County Low Birth Weight Study.' These were 4.5 to 5.1 for term LBW and 2.2 to 3.3 for preterm LBW. Probability of Nonnal Birth Weight (> 2500g) at Tenn (? 37 weeks) for Con- tinuingSmokers (Nodes 4,5,6). The prob- ability of normal birth weight at term for each smoking category was derived by subtraction: Prr«m = 1-Pe~ceRn-PTe., where Matemal Outcomes Probability of Antepartum Hemor- rhage, Abraptio Placenta, and Placenta Previa for Smokers, Nonsmokers, and Quitters (Nodes 7, 8, 9). Studies16-23 re- port elevated relative risks among smok- ers for placental abnormalities or bleeding during pregnancy. We converted these relative risks to probabilities using the for- mula cited earlier. An incidence of .01 for hemorrhagez¢26 and a relative risk of 1.43 for hemorrhage in smokers'=' were used in the baseline formula, and we obtained the probability of hemon; hage in continuing smokers (node 8) by the same method we used for infant outcomes. Probability of Preeclampsia in Smokers, Nonsmokers, and Quitters (Nodes 7, 8, 9). The formula cited earlier was used to calculate the probability of preeclampsia in nonsmokers and quitters. The incidence rate of preeclampsia for pregnant women in the United States (.05) was obtained from studies cited in general obstetrics texts ux The baseline relative risk of preeclampsia for smokers (.42) was Costs Because true cost data cannot be ob- tained and results of this analysis are of most interest to third-party payors, charges were used as a proxy for costs. We included in our analysis only the direct medical charges for maternal care at de- livery and for hospital care for newborns. With no national data available at the level of detail needed for this analysis, we ap- proximated charges based on 1989 hospi- tal and physician charges from two San Francisco Bay Area hospitals with peri- natal databases. Because exact obstetric and newborn charge data are unavailable to compare San Francisco hospital charges with national averages, we varied the ranges of charges in our sensitivity analysis from 50% to 200% of the San Francisco charges, a range that includes most hospital charges in the United States.30 The cost assumptions used in the model are summarized in Table 2 and were derived as described below. 386 American Journal of Public Health TIMN 346472 t4(arch 1992, Vol. 82, No. 3

smofdng Cesvation Programs in eregnancy I Maternal Chaiges. We identified all patients discharged from the two study hospitals with anteparturn hemorrhage or preeclampsia (ICD-9 codes 641.11, 641.21, 642.41, and 642.51)31 between July 1 to December 30, 1989. We took a sys- tematic random sample of women who de- livered during the same period without hemorrhage or preeclampsia. We did not include maternal hospi- talization prior to delivery for any of the diagnoses, given that reliable data were not available on this parameter. However, predelivery charges are unlikely to make a major contribution to total charges. We obtained hospitalization bills for 94% of the women identified (124 of 132), and to each bill we added a standard obstetrician charge for the delivery of an infant. For our mean baseline value for both infant and maternal charges, we elimi- nated the highest and lowest values, pre- suming them to be outliers. Infant Charges. All infants diagnosed as term LBW born in the study period were identified using the obstetrical data- base of each hospital (n = 31). For infants who were preterm LBW and term not LBW, we took a systematic random sam- ple of discharges from the same period (n = 30 for each diagnosis). For all these infants, we obtained hospitalization bills for 96% (87 of 91). We approximated physician charges for preterm and term LBW infants by tak- ing 20% of hospital charges for each in- fant, a number found in an earlier study,32 to be the average physician charge for one of the same Bay Area hospitals used in our study. For the infants who were not LBW, we used a standard charge billed by the pediatric group of each institution for care of a normal newborn. Sensitivity Analysis for Costs. Be- cause hospital and physician charges for the same diagnosis may vary widely due to regional and institutional characteris- tics, we performed a sensitivity analysis by varying the amount charged for each diagnosis from 50% to 200%. Results Table 3 summarizes the estimated break-even cost per pregnant woman of a program for smoking cessation in preg- nancy, based on various outcomes and population characteristics. Using our baseline assumptions and considering only infant outcomes, we estimate that $35 per pregnant woman is the largest amount the program could cost without exceeding the cost of care for LBW infants later on. March 1992, Vol. 82, No. 3 TABLE 2-Homp[ql! and WtyMden MMrt thary.s ior SNecw Dttgnosss Aswdmed tft 8moidtp te PY+ogrtsncy: Ssn Francisco, ig64 Ntmber of Paderft Mesn Charges (S) Mate<ttel f{ernant"e 35 11056 PreaclanVela 51 8891 No hentonhage or preedarM 38 6794 tr>Fart <37 wks, <2500 g 28 43 755 2:37 wks, <25po g 29 4978 a;37 wks, z:2800 g 30 2738 TABLE 3-Zs*rsaftd Bredc-Ewn Cos!' per Pregnant Women of a Propram for Srnok- Itq Cesstllon Nt PreprNnoy, by OuOcoms Considered and Risk Charac- MrbdilCS of Popuiatlon In Teans of LOW Oul3corres Considered (S) Intent I Risk of LOW Oniy Maternal Baseline 35 32 LOW incideriCe preterm LBWb 28 25 tiigh inCicfence "arm LBW° 87 64 t.nuv incidetlce Term LBVW 30 27 High incidence Term LBW' 64 61 'n~ d doAers per prognart wornen thet ooek be irneeted in srnddng cessalJon program w8h a net opmdtre of zeto adbr teklrtg inb aooourR comts due to cherges in cost of medcai car® tor other ~ °Ueir~g~ p iobe bUly e*meN of prelerm I.BN tor nanmoWrs and qurners (.029). Ui+a the uPper probabilitr o.orrft of pelerm lJ3LV for rwramacers er,d quNtera (.066). °Ueirg " lower probetiQty ewmel® d temf !BW for norWnokers ard qu"s (015). 'Uring the upp®r prob®Dilgy eeYmaOS d iarm LOW for norarmkers and quittsrs (.031). This amount is the estimated break-even cost. Using our baseline assumptions and considering both maternal and infant out- comes, the break-even cost of a program for smoking cessation during pregnancy is estimated to be $32 per pregnant woman. This is $3 less than the cost considering only infant outcomes because the higher risk of preeclampsia in nonsmokers re- sults in an increase in expenditures. Considering only infant outcomes, estimates of the break-even cost of such a program range from $28 to $67, depending on the assumed background incidence of preterm LBW in a population, and from $30 to $64, depending on the assumed in- cidence of term LBW (Table 3). When both maternal and infant outcomes are considered, the estimates of break-even cost are not much different. Using the baseline assumptions, es- timates of the break-even cost vary, de- pending on the percentage of women who smoke at the first prenatal visit. These es- timates range from $12 per woman in a population with 10% smokers to $45 per TIMN 346473 US, a'~_. 0 30 10 N Porc.M or Smo•a., at Fu,t V,ot FIGURE3-EstlmeEed break-even cost of a program for smokkV cessation during preg- nancy, by prevalence of smoking at the first prena- tel vWt, cortsldering both kHant and maeemal out- comes and using baseline ataunptlorts for an otlw varfables. woman in a population with 35% smokers (Figure 3). The estimated break-even cost strongly depends on assumptions about American Journal of Public Health 387

Shipp et al. 625 t5 20 25 30 Aswm y fi¢IaMe qnk CI PrMEtm LBW n SmaNUf FIGURE 4-Estimated break-even cost of a program for smoking cessation dur- ing pregnancy for differ- ent assumptions about the relative risk of pre- term LBW and several estimates of the back- ground incidence of pre- term LBW, considering both Infant and matemal outcomes and using baseline assumptions for all other variables. 0 35 40 45 5. AssumaC H.utuv. R,sx a1 rum Lew m Smokua FIGURE 5-Estimated break-even cost of a program for smoking cessation dur- ing pregnancy for differ- ent assumptions about the ntlative risk of term LBW and several esti- mates of the background incidence of preterm LBW, considering both Infant and maternal out comes and using baw line asstxttptlorts for alN other variables. the relative risk of preterm LBW in con- tinuing smokers (Figure 4). If the relative risk of preterm LBW in smokers is 3.0, as has been reported in one study of an urban Black population with a high background incidence of preterm LBW,4 the break- even cost is $215 per woman. The esti- mated break-even cost does not depend on assumptions about the relative risk of term LBW in smokers (Figure 5). TABLE 4-EsiYnabd Bnak-Etan Cost for 9iavM'at AssunWttons about tttt Cost of Variais Oubcames, undsr Bassllnt Assunpttons, ConsidWing Both Mi- tarrtM and intant Ouloornp infant Matemal {V®t Cost (t) Baseline 35 (-3) 32 Preterm L$U1l 50% 19 (-3) 18 Pretemt LBW 200% 67 (-3) 64 Term LBW 50% 30 (-3) 27 Term LBW 200% 46 (-3) 43 Preedampsia 50% 35 (+5) 40 Preedampsia 2 0Q% 35 (-19) 16 ~ ~~ l~en~il.y~afle W%O 35 (-4) 31 Hemonhage 200% 35 (-1) 34 As might be anticipated, assumptions about the effectiveness of the smoking cessation program influence estimates of the break-even cost. If the rate of smoking cessation were as low as 10% as a result of the program, the break-even cost drops to $12 per woman; however, if the program results in a net rate of smoking cessation as high as 29%, the break-even cost is $41. Assumptions about the cost of care for a preterm LBW infant affect the esti- mate of the break-even cost more than assumptions about the cost of medical care for term LBW infants or for either of the maternal complications whose risk is increased in smokers (Table 4). No rea- sonable estimate of the cost of medical care for preeclampsia results in a net in- crease in expenditures when both infant and maternal outcomes are considered. Discussion Our analysis shows that, even when considering only direct medical charges related to hospitalization at delivery, a smoking cessation program that invests $32 for every pregnant woman seen in the facility would still break even economi- cally in the short term. The number of dollars that could be invested in a smoking cessation program and still allow the pro- gram to break even is greater for popula- tions with higher incidences of LBW, greater relative risks of LBW in smokers, and higher prevalence of smoking than the national average. However, at every level of LBW and of smoking prevalence found in the literature, money could be invested in a smoking cessation program and the program would still break even. It should be noted that, consistent with the program perspective of our anal- ysis descnbed earlier, our "per patient" cost estimates apply to a cost for all pa- tients, combining smokers and nonsmok- ers. To obtain the amount that could be invested per smoker, the "per patient" cost is divided by the prevalence of smok- ing in the population (e.g., in a baseline population with a smoking prevalence of 25%, $32/.25, or $128.00, could be spent for each smoker). The largest cost associated with med- ical care of pregnant smokers and their infants is for infant care and not for care of maternal complications. There are three reasons for this. First, adverse infant out- comes are more common in smokers than are adverse maternal outcomes. Second, charges for adverse infant outcomes are much higher than those for maternal com- plications. Third, preeclampsia is less common in smokers than in nonsmokers. The estimate of a break-even cost was most sensitive to assumptions about the incidence of pretenm. LBW and the relative risk for preterm LBW for smok- ers. For instance, in a population studied recently with a high incidence of and high relative risk for preterm LBW, the break- even cost was $237.4 Although the prev- alence of smoking at the first prenatal visit and the success rate of the formal smoking cessation program also affected the cost analysis, within the range of probabilities in the literature they did not produce as much variability in the break-even cost. The break-even cost is moderately influenced by the quit rate for smokers attributable to the program, a rate that has varied considerably for programs studied.&4.12 Our baseline analysis is de- rived from a fairly optimistic quit rate of 23% obtained from an intensive interven- tion program for pregnant women that in- cluded one home visit, one telephone call per month, and twice-per-month mailings of smoking cessation literature.7 More modest quit rates of 12% to 13.6% were 388 American Journal of Public Health TIMN 346474 March 1992, Vol. 82, No. 3

obtained by programs that included only short health education sessions and printed materials 6•s These more modest programs cost only $7.13 to $11 per preg- nant smoker,8,33 an amount below our cal- culated break-even cost when we substi- tuted these quit rates into our analysis. There are several important limita- tions to our study. First, we included only medical charges for the hospital admission for delivery and in-hospital charges for the neonate after delivery; therefore, our es- timates are conservative. Second, we used local charge data because they are precise for LBW diagnoses. However, in our sensitivity analysis, we varied charges for each diagnosis from 50% to 200% of the mean charge, figures that include the charges for services at most US hospi- tals.30 Last, we included only direct med- ical costs in our analysis. Cost analyses of other conditions often include indirect costs to society associated with years of potential life lost per premature death34-35 and days of work lost due to illness.34-36 Moreover, some studies have shown in- creased infant mortality in babies of smok- ers 37•38 Still others suggest that children of smokers may have deficiencies in growth, intellectual development, and be- havior39-41 or an increased risk of otitis media.42,43 Again, not considering these other adverse effects of smoking in preg- nancy makes our analysis conservative. A recent cost-effectiveness analysis was done by Marks et al.,9 who hypoth- esized that $30 per pregnant smoker could be spent for a model smoking cessation program; with a 15% cessation rate, this would cost $23 505 300 nationally per year, or $4000 per LBW baby averted. The study compared this figure with NICU costs for LBW babies of smokers and surmised that the program would save $3.31 for every $1 invested. This ratio would increase to over $6 saved for every $1 spent if the costs of long-term care for infants with disabilities secondary to LBW were included. Although this study differed from the present analysis in sev- eral ways, the major difference was one of perspective: their analysis calculated the costs for a smoking cessation program na- tionally and compared them with the costs of care for LBW infants; our study takes a program perspective. Importantly, our analysis includes a way of determining which variables the analysis is most sensitive to, and it pro- vides a way for planners to input their own values for these population variables. Both our study and the Marks et al. study, as well as the HMO study mentioned ear- lier,10 conclude that smoking cessation programs in pregnancy are cost-effective, whether from a national or a program per- spective. Our analysis adds a dimension that allows program planners to determine how much to spend on these programs. Program planners and administrators can use our analytic framework along with their own data to help decide how much to invest in a smoking cessation program. However, planners may want to consider factors other than the immediate break- even cost in their decision to include a smoking cessation program in their pre- natal care. Considering the other long- range adverse medical, psychological, and societal outcomes of smoking, planners and society should be willing to invest in smoking cessation programs for pregnant women, regardless of their direct benefit of saving money. 0 Acknowledgments This work was supported in part by a fellowship grant from the Pew Charitable Trusts and funds provided by the cigarette and tobacco surtax fund of the State of California through the To- bacco-Related Disease Research Program of the University of California. The authors also acknowledge Marcus McCrory and the em- ployees of San Francisco General Hospital for their invaluable assistance with the charge data. References 1. Athayd E. Incidencia de abortos e morti- nalidade nasoperarias da industria de fumo. Btsil-Med 1948;62:237-239. 2. Simpson WJ, Linda L. A preliminary re- port on cigarette smoking and the incidence of prematurity. Am J Obstet Gynecol. 1957;73:808-815. 3. Kramer MS. Determinants of low birth weight: methodological assessment and metaanalysis. Bulletin of the World Health Otganization. 1987;65:663-737. 4. Alameda County Low Birth Weight Study Group. Cigarette smoking and the risk of low birth weight: a comparison in Black and White women. EpfdemloL 1990;1:201- 205. 5. Ershoff DH, Aaronson NK, Danaher BG, Wasserman FW. Behavioral, health, and cost outcomes of an HMO-based prenatal health education program. Public Health Rep. 1983;98:536-547. 6. Windsor RA, Cutter G, Morris J, et al. The effectiveness of smoking cessation meth- ods for smokers in public health maternity clinics: a randomized trial. Am J Public Health. 1985;75:1389-1392. 7. Sexton M, Hebel JR. A clinical trial of change in maternal smoking and its effect on birth weight. JAMA. 1984;251:911-915. 8. Ershoff DH, Mullen PD, Quinn VP. A ran- domized trial of a serialized self-help smok- ing cessation program for pregnant women in an HMO. Am J Public Health. 1989;79:182-187. 9. Marks JS, Koplan JP, Hogue CJR, Dalmat ME. A cost-benefit/cost-effectiveness anal- smoking Cessation Programs in Pregoancy ysis of smoking cessation for pregnant women. Am J Pn?v Med. 1990;6(5):282- 289. 10. Ershoff D, Quinn VP, Mullen PD, Lairson DR. Pregnancy and medical cost outcomes of a self-help prenatal smoking cessation program in a HMO. Public Health Rep. 1990;105(4):340-347. 11. National Center for Health Statistics. Y'ual and Health Statistics. Series 10, No. 163. 1988. DHHS publication PHS 88-1591. 12. Multiple Risk Factor Intervention Trial Re- search Group. Multiple risk factor inter- vention trial: risk factor changes and mor- tality results. JAMA. 1982;248:1465-1477. 13. MacArthur C, Newton JR, Knox EG. Ef- fect of anti-smoking health education on infant size at birth: a randomized controlled trial. Br J Obstet GynaecoL 1987;94:295- 300. 14. Puffer RR, Serrano CV. Patterns of Birth Weight. Washington, DC: Pan American Health Organization, World Health Orga- ni7ation; 1987. Scientific Publication 504. 15. Ounsted M, Moar VA, Scott A. Risk fac- tors associated with small-for-dates and large-for-dates infants. BrJ Obstet Gynae- coL 1985;92:226-232. 16. McIntosh ID. Smoking and pregnancy: at- tributable risks and public health implica- tions. Can J Public Health. 1984;75:141- 148. 17. Lowe CR. Effect of mothers' smoking hab- its on birth weight of their children. BrMed J. October 1959;10:673-676. 18. Underwood P, Hester LL, Luffitte T, Greg KV. The relationship of smoking to the out- come of pregnancy. Am J Obstet GynecoL 1965;91:270-276. 19. Underwood PB, Kesler KF, O'Lane JM, Callagan DA. Parental smoking empirically related to pregnancy outcome. Am J Ob- stet GynecoL 1967;29:1-8. 20. Butler WR, Albermon ED. Perinatal Prob- lems: The Second Report of the 1958 Brit- ish Pennatal Mortality Survey. Edinburgh and London: Livingstone; 1969. 21. Andrews J, McGany JM. A community study of smoking in pregnancy. J Obstet Gynaecol Br Common w. 1972; 79:105 7- 1073. 22. Lubs ME. Racial differences in maternal smoking effects on the newbom infants. Am J Obstet GynecoL 1973;115:66-76. 23. Meyer MB, Jonas BS, Tonascia JA. Peri- natal events associated with maternal smoking during pregnancy. Am J Epide- miol. 1976;103:464476. 24. Cunningham GF, MacDonald PC, Gant NF. Williams Obstetrics. 18th ed. Nor- walk, Conn: Appleton and Lange; 1989. 25. Reid DE, Ryan KJ, Benirschke K. Princi- ples and Management of Human Repro- duction. Philadelphia, Pa: WB Saunders Co; 1972. 26. Goplerud CP. Bleeding in late pregnancy. In: Danforth WJ, ed. Obstetrics and Gyne- cology. 4th ed. Philadelphia, Pa: Harper and Row; 1982:443-456. 27. Marcoux S, Brison J, Fabia J. The effect of cigarette smoking on the risk of pre-ec- lampsia and gestational hypertension. Am J Epldemiol 1989;130:950-957. 28. Ylikorkala 0, Makila UM. Prostacyclin March 1992, Vol. 82, No. 3 American Journal of Public Health 389 TIMN 346475

Shipp et al. and thromboxane in gynecology and ob- stetrics. Am J Obstet GynecoL 1985;52: 318-329. 29. Ylikorkala 0, Vilnikka L, Lehtovirta P. Ef- fect of nicotine on fetal prostacyclin and thromboxane in humans. Obstet GynecoL 1985;66:102-105. 30. American Hospital Association. Hospital Statistics. Chicago: American Hospital As- sociation, 1990. 31. International Classification of Diseases-9- Clinical Modification, Volume 1, Anno- tated Edition Fourth Printing. Ann Arbor, Mich: Edwards Brothers; 1988:537-541. 32. Phibbs LS, Williams RL, Phibbs RH. New- born risk factors and costs of neonatal in- tensive care. Pediatrics. 1981;68:313-321. 33. Windsor RA, Warner KE, Cutter GR. A cost-effectiveness analysis of self-help smoking cessation methods for pregnant women. Pub Health Rep. 1988;103:83-88. 34. Rice DP, Hodgson TA, Sinsheimer P, Browner W, Kopstein AN. The economic costs of the health effects of smoking, 1984. Milbank Q. 1986;64:489-547. 35. Morbidity and Mortality Weekly Report. Smoking-attributable mortality, morbidity, and economic costs: California, 1985. JAMA. 1989;261:2942-2945. 36. Swank RT, Becker DM, Jackson CA. The costs of employee smoking: computer sim- ulation of hospital nurses. Arch Intern Med. 1988;148:445-448. 37. Malloy MH, Kleinman JC, Land GH, Schramm WF. The association of maternal smoking with age and cause of infant death. Am J EpidemioL 1988;128:46-55. 38. Kleinman JC, Pierre MB, Madons JH, Land GH, Schramm WF. The effects of maternal smoking on fetal and infant mor- tality. Am J EpidemioL 1988;127:274-282. 39. Butler R, Goldstein H. Smoking in preg- nancy and subsequent child development. Br Med J. 1973;4:573-575. 40. Dunn HG, McBurney AK, Ingram S, Hunter CM. Maternal cigarette smoking during pregnancy and the child's subse- quent development: I. physical growth to the age of 6V2 years. Can J Public Health. 1976;67:499-505. 41. US Dept of Health and Human Services. The Health Consequences of Smoldng for Women: A Report of the Surgeon GeneraL 1983:191-244. DHHS publication 410-a9/ 1284. Reed BD, Lutz LJ. Household smoking ex- posure: association with middle ear infec- tions. Fam Med. 1988;20:426-430. 42. 43. Hinton AE, Buckley G. Parental smoking and middle ear effusions in children. J Laryngol OtoL 1988;102:992-996. Finding Fault before Drillx'ng: A Sot.ution for Avokding Radon Dr. Itina Cech calls it a $2 solution. It's a solution, how- ever, that could add up to a great deal of financial and health savrngs, not to mention saving headaches, particularly for those concerned with developing public water supplies. Cech, professor of environmental health sciences and hy- drology at the Sctmol of Public Health of The Utuversity of Texas Health Science Center at Houston, has some simple advice: to avoid a radon-contaminated well, don't drill within 4 miles from a salt dome, fault, or areas known or suspected to have uranium deposits. Radon is a radioactive byproduct of uranium decay. Be- cause it is odorless and colorless, there is nothing to alert well drillers or household residents unless testing occurs, Cech. said. "It is very costly to drill a well only to find out that the radonconcentratiatis too high. Byexamining the area for fault tines before drilling, the problem can be avoided," she said. Cech said uranium dept3sit.s appear primimly along the coastline of Texas, but dot other areas of the state as wen. Radon deposits also collect along the flaalCs of aait domes, also common to the coastai areas.. She aaici thuits in the domes allow radon to "leak" ttfawsuntL If an aquifer, or undesgrotud layer of earth saturated with water, happens to fie above or neara satt dome, radoctmayenterthewatersuppEythmugh the faults. - - Todate, Cechand herstudents havegathered almost 1000 water samples in Texas from public and individually owned domestic water wells and surface sources, along with wella used for livestock watering, industtlsl purposes, and od pro- duction. Although radon can escape through the soil and seep tlmough foundation cracks, in Texas it isprunarityfonrtd in the water supply, Ceeh explained. "Concentrations of radon were p®rtieularly high in public water wells along the Gulf Coast and in parts of west-central Texas," she said, explaining that the same ares has one of the highest rates for respiratory cancer mortality in the nation. . "What we found was very interesting. When wells are developed around faults, a greater likelihood exists for en- countering radon," she saic}. The maximum acceptable limit of radon concentration in public drinking water that the Environmental Protection Agency has proposed is 300 pCi/L. According to the EPA, the lifetime risk of ckvebping lung cancer from household water that contains 10 000 pCy/L of radon is roughly 3 to 13 in 1000. "About 503'0 of the public water supplies measured byour group exceed the recommended standards. The highest cocr centration, 22 000 pCS/L, was found in southwest Texas near Refugio. In Houston, the highest observed was 6000 pC'tfL," she said. "I tva is of concern but should not be cause for alarm since radon levels are correctable," she said, explaining that public utilities will have until 1996 to meet the EPA rtquirer menta. Aeratiug the water suppiy would allow radon to escape before the water is piped to homes and businesses. Radon in water is a problem only when the chemical is released fram the water and enters household air. As one breathes, the radon breaks dcxvu further, causing small bursts of energy that can damage hutg tissue. The chemical is re- leased, for exampk, when people wash dishes and clothea6 take sbowers, and Rt>stt ta'fetr, "Radon is more likely to be present in homes that ara heavily insulated and airt,"ght, which a extremely common in Texas. Drinking the water is not considerai as much of a hazard as bc+eathalg the radori," she aaid. Economical ways to reduce risit fmm rsdon are to hicrease air floov in and out aEt6e house and to retrain fooaa smoking (scientific evide=we itndi- cates that sntobag may increase the riatc o(eapoMtre to rrdon}. Because radon appears primarily along the coastal bor- ders, it is bdie%vd that asinngar aidiations exist in other puts of the world« TIMN 346476 I 390 American Journal of Public Health March 1992, Vol. 82, No. 3