[Skip to Navigation]
Sign In
July 2004

Secondhand Smoke Exposure, Smoking Hygiene, and Hospitalization in the First 18 Months of Life

Author Affiliations

From the Department of Community Medicine and School of Public Health, University of Hong Kong, Hong Kong, China.

Arch Pediatr Adolesc Med. 2004;158(7):687-693. doi:10.1001/archpedi.158.7.687

Background  Recognizing the suboptimal public health effects of a complete cessation strategy for parents and child caregivers who smoke, some researchers have called for a harm reduction approach; however, the evidence remains scanty and controversial.

Objective  To examine the effects of secondhand smoke and smoking hygiene on infant health and related health care costs during the first 18 months of life.

Methods  We conducted prospective follow-up on 8327 newborns during April and May of 1997 for 18 months in a population-based birth cohort of infants from Hong Kong, China.

Main Outcome Measures  Number of hospital admissions, adjusted odds ratios for ever hospitalization for each secondhand smoke exposure variable, and corresponding population attributable risks.

Results  Most secondhand smoke exposure came from fathers and other household contacts, whereas only 2.8% of mothers smoked postpartum. The odds ratio of ever hospitalization due to all illnesses combined for infants living in a household with any (maternal, paternal, or other) smoker who was smoking at least 3 m away from the infant, as reported by a parent, was 1.00 (95% confidence interval, 0.88-1.13) compared with those in a smoke-free household. The corresponding odds ratio for infants living with any smoker at home with poor smoking hygiene (<3 m away) was 1.28 (95% confidence interval, 1.07-1.52), which translated into 2.8% of all inpatient episodes in the first year of life, representing an additional 616 admissions.

Conclusions  Hospital admission was significantly more likely among infants exposed to secondhand smoke if it was accompanied by poor smoking hygiene. Harm reduction strategies should be rigorously adhered to when complete cessation is not possible.

Three recent systematic reviews1-3 of a large and accumulating body of scientific evidence to date have conclusively confirmed the harmful effects of secondhand smoke on infant and child health. Parental smoking is directly responsible for higher rates of both upper and lower respiratory tract illnesses including otitis media, asthma, bronchitis, and pneumonia. Therefore, the UK Royal College of Paediatrics and Child Health's National Service Framework for Children, Young People and Maternity Services4 advocates the progressive reduction of population smoking prevalence in mothers and other high-risk groups such as manual laborers or the lowest socioeconomic stratum. In the United States, the American Academy of Pediatrics calls on all health care professionals to routinely inquire about household smoking history and offer smoking cessation services to parents who smoke.5

Sustained quitting in most adult household members and child caregivers who smoke is difficult to achieve. A recent Cochrane review of randomized controlled trials reported that overall only 17% (range, 14%-24%) of smokers who were offered nicotine replacement therapy maintained abstinence at 12 months.6 This rate can be improved to about 30% with additional individual behavioral counseling7 or telephone counseling with repeated callback and follow-up.8 Thus, 7 of 10 smokers who are willing to attempt to stop smoking are not able to quit despite current pharmaceutical aids and psychobehavioral interventions. This does not include the 80% of smokers who are not planning to quit within the next 6 months.9 Recognizing the suboptimal health effects of a complete cessation strategy, some researchers have called for an evidence-based harm reduction approach.10

In the case of secondhand smoke and child health, the promotion of proper smoking hygiene (ie, smoking away from the immediate environment of infants and children) has been recommended by the American Academy of Pediatrics.5 This guidance was partly based on studies reporting indirect evidence suggesting a negative correlation between the distance a parent smokes away from the infant and that infant's urinary cotinine level11,12 and that a lower cotinine level may be associated with a reduced risk of respiratory tract infections.13,14 Nevertheless, harm reduction through improved smoking hygiene has yet to be comprehensively and scientifically validated in clinical and population settings with relevant patient-oriented and public health outcomes such as illness incidence and service utilization.

In this article, we report a population-based birth cohort study in Hong Kong, China, in which our objective was to study the effects of secondhand smoke and smoking hygiene on infant health, use of inpatient health care, and related direct medical costs during the first 18 months of life. The null hypothesis for the study was that the proximity of smoking (arbitrarily dichotomized as >3 m vs <3 m from the infant) makes no difference in hospitalization rates in the first 18 months of life.


Data sources

Details of the study methods have been published previously,15,16 a summary of which follows. Data for this analysis were obtained from a prospective birth cohort study conducted at all 47 Maternal and Child Health Centres (MCHCs) of the Hong Kong government's Department of Health in 1997. Most infants in Hong Kong, regardless of the family's ability to pay, visit the MCHCs for preventive care and immunizations. For the index year, 92% of all infants born in Hong Kong visited an MCHC at least once. The sampling frame consisted of all infants brought to an MCHC for their first visit after birth in April and May 1997. The response rate to our study was 95%, accounting for 88% of all births during that period. There were 8327 mother-infant pairs in the final cohort.

Mothers provided information on breastfeeding history, household smoking habits, mode of delivery, and other demographic, obstetric, behavioral, and potentially confounding variables via a standardized self-administered questionnaire at their first MCHC visit (baseline) and at 3, 9, and 18 months after birth. Questions ascertaining household smoking patterns (eg, "Did the infant's mother smoke in the previous week?") included maternal and paternal smoking as well as smoking by other household members, both during pregnancy and after birth. If a household member smoked, respondents were asked if this person usually smoked within or beyond 3 m of the infant. Utilization of inpatient health services was measured by the number of hospital admissions since the last MCHC follow-up visit. Mothers were also requested to provide information on the type of illness (respiratory tract illness, febrile illness, or other) leading to the hospitalization. To minimize recall bias, parents were given health services utilization questionnaires to take home and were asked to fill out the form whenever the infant was hospitalized. The outcomes of interest were prospectively documented and recorded as they occurred. Using telephone interviews, trained research assistants contacted subjects (<5% mainly owing to illiteracy) who could not complete the questionnaire and those who did not return it to ensure optimal follow-up. The project received approval from the Ethics Committee of the Faculty of Medicine, University of Hong Kong.

Statistical analysis

We used bivariable and multivariable logistic regression analyses to study the association between secondhand smoking patterns (taking into account smoking habits of household members and proximity to the infant when smoking, as reported by a parent) and hospitalizations of infants, as recorded during follow-up visits at ages 3, 9, and 18 months. Secondhand smoking patterns as recorded on the baseline survey were used as the primary predictor variable. Hospital admissions were dichotomized as ever vs never utilization. Variables in the multivariable model were selected if they were associated with P<.05 in bivariable analyses or were included based on known confounders documented in the previous literature.1,2 The following independent variables were included in the final model: breastfeeding practice (ever vs never), birth weight (<2500 g, 2500-2999 g, 3000-3499 g, 3500-3999 g, or ≥4000 g), method of delivery (normal vaginal, vaginal assisted with suction or forceps, or cesarean birth), mother's age (≤24, 25-29, 30-34, or ≥35 years), mother's education level (≤6th grade, 7th-9th grade, 10th-11th grade, or matriculation/university), full-time job held by mother (yes or no), and birth order of infant (first, second, third, or later).

To assess the incremental adverse health effects of secondhand smoke, taking advantage of the population-based design of the study, we calculated the respective adjusted population attributable risks and their 95% confidence intervals (CIs) directly from logistic regression for each of the passive smoking exposure variables.17 This statistical approach allows for an unbiased estimate of the population attributable risk even when the prevalence of the outcome of interest (ever utilization in our case) does not satisfy the assumption of rare disease. It also adjusts the estimated effect of one form of secondhand smoke exposure for the effects of others and potential confounders. Only variables shown to be significantly associated with a greater use of health services in the earlier part of the analysis were included in population attributable risk calculations. All analyses were conducted using Stata version 8.0 statistical software (Stata Corp, College Station, Tex).


Overall, 28.6% of the 1997 birth cohort of infants from Hong Kong were admitted at least once for inpatient care during the first 18 months of life, about two thirds of whom (18.3% / 28.6% = 64.0%) had respiratory tract or febrile illnesses. For those who were ever hospitalized, the mean ± SD annual number of inpatient episodes was 1.3 ± 1.1 overall and 1.2 ± 0.9 for respiratory tract or febrile illnesses (Table 1).

Table 1. 
Distribution of Infant Characteristics and Incidence of Hospital Admission*
Distribution of Infant Characteristics and Incidence of Hospital Admission*

Table 1 also presents the distributions of infant characteristics. Most mothers (69.4%) were aged 25 to 34 years. Six (59.7%) of 10 had continued beyond the minimum 9-year education required in Hong Kong, and slightly fewer than half (46.2%) worked full-time during the baseline visit (excluding maternity leave). More than one quarter (26.9%) of the infants were delivered by cesarean birth, and an additional 16.6% had instrument-assisted vaginal birth. Most infants were the first or second child. Their birth weights were normally distributed. Of the mothers, 45.0% had initiated breastfeeding for a median of 8 weeks (data not shown). All of these covariables except for birth order and maternal full-time employment showed statistically significant associations with hospital utilization in bivariable analyses.

In terms of household smoking patterns, 4.6% of mothers smoked during pregnancy; however, the proportion shrank to 2.8% postpartum, when most (2.2% / 2.8% = 78.6%) were reported to have smoked at least 3 m away from the infant. Slightly more than one quarter of fathers reportedly smoked outside a 3-m radius of the infant, and an additional 7.1% smoked within 3 m. The corresponding proportions for other household contacts, mostly grandparents, were 10.9% and 4.4%, respectively. Overall, 41.2% of infants in this birth cohort were exposed to secondhand smoke at home, where one quarter (10.4% / 41.2% = 25.2%) of smokers in the household smoked within 3 m of the infant. These statistics are consistent with results from previous surveys of children and youth in Hong Kong during the last decade18-20 (Table 2).

Table 2. 
Incidence of Ever Hospital Admission by Smoking Variables and Associated Adjusted Odds Ratios*
Incidence of Ever Hospital Admission by Smoking Variables and Associated Adjusted Odds Ratios*

Table 2 shows the parent-reported incidence and adjusted odds ratios (ORs) of ever hospitalization stratified by exposure to different forms of secondhand smoke in the household and by diagnostic group (respiratory tract or febrile illness or any illness). Hospital admission due to respiratory tract or febrile illness as well as for any illness was significantly more likely among infants exposed to secondhand smoke if it was accompanied by poor smoking hygiene (ie, parental-reported smoking within 3 m of the infant) for all postpartum smoking exposure categories. Statistical significance was attained only for paternal smoking and the overall category of any smoker at home. This was likely because of the low maternal smoking prevalence and relatively less intensive contact between nonparental household members and infants. There was a consistency of effect across all secondhand smoke categories after birth in which parent-reported smoking at least 3 m away from the infant yielded nonsignificant ORs, most of which approximated unity (eg, OR of ever hospitalization due to any illness for any smoker at home but not within 3 m, 1.00; 95% CI, 0.88-1.13). In contrast, we observed relative odds increases of 27% (P = .02) owing to respiratory tract or febrile illness and 28% (P = .006) for all illnesses combined in infants living in a household with any (maternal, paternal, or other) smoker.

During the in utero period, infant exposure to secondhand smoke was through the mother's environment. There was a clear dose-response gradient (P = .008) from no exposure to passive smoking and active smoking, in which the corresponding ORs associated with ever hospitalization were 1.00, 1.15 (95% CI, 1.02-1.29), and 1.33 (95% CI, 1.02-1.73). Similar results were obtained for respiratory tract or febrile illness alone except that secondhand smoke exposure did not reach significance at P<.05 (Table 2).

Table 3 details the population attributable risks due to secondhand smoke for associations that achieved statistical significance, as reported previously. For our 1997 annual birth cohort, maternal passive and active smoking during pregnancy were responsible for generating an additional 1858 and 326 inpatient episodes, respectively, in the first year of life. Similarly, postnatal secondhand smoke exposure at home within 3 m of the infant from parents and other household contacts led to 616 extra hospitalizations in this population.

Table 3. 
Population Attributable Risks of Hospital Admissions in the First Year of Life Owing to Poor Household Smoking Hygiene*
Population Attributable Risks of Hospital Admissions in the First Year of Life Owing to Poor Household Smoking Hygiene*


Our findings provide direct evidence from a postindustrialized population in China that secondhand smoke exposure at home, if accompanied by poor smoking hygiene (ie, smoking within 3 m of the infant) as reported by a parent, was positively associated with higher hospital admission rates due to respiratory tract or febrile illness and for all illnesses combined in the first 18 months of life. In addition, infant exposure to secondhand smoke via mothers' active and passive smoking during pregnancy contributed to more frequent inpatient utilization; the magnitude of effect was at least that of postnatal exposure via household contacts. These results have important public health effects in terms of increased health service utilization. More than one tenth of inpatient care in the first year of life can be attributed to secondhand smoke exposure before and after birth and could be partially averted if proper smoking hygiene practices were universally adopted (in this case, if household smokers stayed at least 3 m away from the infant as reported in a prospective parent-completed questionnaire).

Most secondhand smoke exposure came from fathers and other household contacts, whereas only 4.6% and 2.8% of mothers, respectively, smoked during pregnancy and postpartum. It is clear from these results that paternal smoking was the main factor in the increased health risks observed. This shows that even in a society with highly effective tobacco control in terms of maternal smoking, the effects of tobacco consumption and related smoking hygiene of fathers on infant health can be substantial. Although not significant at P<.05 because of the small number of maternal smokers, one would suspect that maternal smoking would demonstrate similar findings to those of paternal smoking, as indicated by the close resemblance in the point estimates and dose-response gradients between the 2 sets of results in Table 2.

Our results confirm previous indirect evidence examining harm reduction strategies using infant urinary cotinine-creatinine ratio or indoor air pollution measurements as the main outcomes of interest,11-14,21,22 translating such biological or laboratory proxies into an actual illness burden with significant public health effects. Moreover, most of the existing literature is disease specific, focusing on conditions such as asthma,21 respiratory tract infections,23 or sudden infant death syndrome.24 In contrast, our study examined respiratory tract illness as well as all causes of morbidity. Finally, the community-based recruitment strategy of this study resulted in a representative population sample, which to our knowledge is the first such set of results in the literature.

Two studies recently suggested that proper smoking hygiene can reduce secondhand smoke exposure in infants.23,25 Blackburn et al25 reported in a cross-sectional study that a complete ban of tobacco in the home was associated with a small but significant reduction in urinary cotinine excretion, whereas less strict harm reduction measures had no detectable effect. In contrast, our prospective findings indicate that smoking outside a 3-m radius of the infant was sufficient to produce an almost 30% reduction in inpatient illness burden. Interpreting these results in the context of Hong Kong, which is one of the most densely populated cities in the world and where most families live in an apartment of only 500 to 1000 sq ft, we expect that harm reduction strategies short of a complete smoking ban in the household can substantially reduce secondhand smoke–related diseases in infancy. However, we are limited by the lack of detailed data further classifying the types of harm reduction activities that parents in the cohort undertook to minimize infant exposure (eg, Did smoking 3 m away mean smoking in a different room or outside the home altogether? Did parents take any additional preventive measures such as opening windows and regularly airing parts of the home where they usually smoked?). An analysis of the Tasmanian Infant Health Survey23 cohort revealed that the risk of hospitalization owing to respiratory tract infection was reduced by 56%, 73%, and 95%, respectively, if the mother smoked in a different room and refrained from smoking while holding and feeding the infant. Our results are in agreement with these findings and extend their application to fathers and other household smokers as well as for respiratory conditions overall and all illnesses combined in a metropolitan Chinese population.

Whether harm reduction strategies can easily be realized in practice remains an important question. A recent Cochrane review26 of family and caregiver smoking control programs for reducing children's exposure to secondhand smoke concluded that brief counseling interventions, successful in the adult health setting when they come from physicians, cannot be extrapolated to adults in the setting of child health. On the other hand, there is limited support for more intensive counseling interventions. Whereas only 4 of the 18 studies reviewed demonstrated a statistically significant intervention effect, 3 of these successful studies used intensive counseling interventions targeted at smoking parents. There is also no clear evidence for differences between the respiratory tract illness, nonrespiratory tract illness, well child, and peripartum settings as contexts for reduction of children's secondhand smoke exposure.

We considered hospitalization but not outpatient consultations because of the often discretionary nature of ambulatory visits, especially in smokers. This phenomenon has been well documented in the previous literature. For instance, 2 large-scale studies in the United States suggested that higher rates of morbidity among children exposed to secondhand smoke might not necessarily be reflected in the higher rates of outpatient consultations.27,28 Several possibilities account for this. First, as White et al29 pointed out in their classic article on the ecology of medical care, only a fraction of those with symptoms go to the physician's office. This iceberg concept of disease is well recognized and provides an explanation for the apparent disconnect between self-reported symptoms and health service utilization. Second, parents are the gatekeepers for their children's access to health care services, and smoking parents often use fewer health care services for their children compared with nonsmokers given the same set of symptoms. It is believed that adult smokers' lower preventive orientation contributes to their decreased use of discretionary health services.28,30 This may translate into underutilization of services for their children as well, especially for non–life-threatening conditions or less serious outpatient consultations. In view of these concerns about the proxy validity of outpatient visits, we focused on the much less discretionary category of inpatient admissions, which would be a more accurate reflection of actual health care need and illness burden. However, whether our findings can be directly extrapolated to the ambulatory setting remains to be investigated. Finally, although we controlled for the potential effects of socioeconomic status by including maternal education in the multivariable model, we cannot rule out residual confounding in other dimensions of socioeconomics such as income and wealth.

In conclusion, if the associations regarding secondhand smoke, smoking hygiene, and illness burden are causal and confirmed in future studies, our findings provide evidence-based support for current recommendations for the promotion of smoking cessation,4,5 or at least proper smoking hygiene practices for those unable to quit despite their best efforts and repeated attempts,5 in all household members. These guidelines should be followed during as well as after pregnancy to reduce both personal and public health hazards due to infant illnesses. Research efforts in identifying effective ways to encourage parents to quit smoking or, if complete cessation is not possible, to practice appropriate harm reduction strategies should be redoubled to translate these findings into a lower illness burden for infants and children currently living in households exposed to tobacco smoke.

What This Study Adds

Recognizing the suboptimal public health effects of a complete cessation strategy for parents and child caregivers who smoke, some researchers have called for a harm reduction approach; however, the evidence remains scanty and controversial.

We conducted prospective follow-up in a population-based cohort of 8327 infants for 18 months in 1997. Our findings indicate that hospital admission was significantly more likely among infants exposed to secondhand smoke if it was accompanied by poor smoking hygiene (ie, smoking <3 m from the infant). Harm reduction strategies should be rigorously adhered to when complete cessation is not possible.

Accepted for publication January 14, 2004.

This study was suported by grant 216106 from the Hong Kong Health Care and Promotion Fund, Hong Kong, China.

We thank the late Connie O, MD, for coordinating the project and all the field work; the Department of Health, Hong Kong, China, for its collaboration and cooperation during data collection; and Marie Chi for expert secretarial assistance in the preparation of this manuscript.

Correspondence: Lai-Ming Ho, PhD, Department of Community Medicine, 21 Sassoon Rd, Faculty of Medicine Bldg, University of Hong Kong, Pokfulam, Hong Kong, China (lmho@hkucc.hku.hk).

DiFranza  JRLew  RA Morbidity and mortality in children associated with the use of tobacco products by other people.  Pediatrics. 1996;97560- 568PubMedGoogle Scholar
Cook  DGStrachan  DP Summary of effects of parental smoking on the respiratory health of children and implications for research.  Thorax. 1999;54357- 366PubMedGoogle ScholarCrossref
Dunn  AZeise  L Health Effects of Exposure to Environmental Tobacco Smoke.  Sacramento California Environmental Protection Agency1997;
Royal College of Paediatrics and Child Health, National Service Framework for Children, Young People and Maternity Services., Not Available Available at:http://www.rcpch.ac.uk/publications/recent_publications/NSF%20Summary.pdfAccessed November 8, 2003
American Academy of Pediatrics Committee on Substance Abuse, Tobacco's toll: implications for the pediatrician.  Pediatrics. 2001;107794- 798PubMedGoogle Scholar
Silagy  CLancaster  TStead  LMant  DFowler  G Nicotine replacement therapy for smoking cessation [Cochrane review].  Cochrane Database Syst Rev. 2002;4CD000146PubMedGoogle Scholar
Lancaster  TStead  LF Individual behavioural counselling for smoking cessation [Cochrane review].  Cochrane Database Syst Rev. 2002;3CD001292PubMedGoogle Scholar
Stead  LFLancaster  TPerera  R Telephone counselling for smoking cessation [Cochrane review].  Cochrane Database Syst Rev. 2003;1CD002850PubMedGoogle Scholar
Etter  JFPerneger  TVRonchi  A Distributions of smokers by stage: international comparison and association with smoking prevalence.  Prev Med. 1997;26580- 585PubMedGoogle ScholarCrossref
Hughes  JR Harm-reduction approaches to smoking: the need for data.  Am J Prev Med. 1998;1578- 79PubMedGoogle ScholarCrossref
Woodward  AGrgurinovich  NRyan  P Breast feeding and smoking hygiene: major influences on cotinine in urine of smokers' infants.  J Epidemiol Community Health. 1986;40309- 315PubMedGoogle ScholarCrossref
Irvine  LCrombie  IKClark  RA  et al.  What determines the level of passive smoking in children with asthma?  Thorax. 1997;52766- 769PubMedGoogle ScholarCrossref
Bakoula  CGKafritsa  YJKavadias  GD  et al.  Objective passive-smoking indicators and respiratory morbidity in young children.  Lancet. 1995;346280- 281PubMedGoogle ScholarCrossref
Gurkan  FKiral  ADagli  EKarakoc  F The effect of passive smoking on the development of respiratory syncytial virus bronchiolitis.  Eur J Epidemiol. 2000;16465- 468PubMedGoogle ScholarCrossref
Leung  GMHo  LMLam  TH Maternal, paternal and environmental tobacco smoking and breastfeeding.  Paediatr Perinat Epidemiol. 2002;16236- 245PubMedGoogle ScholarCrossref
Leung  GMLam  THHo  LM Breast-feeding and its relation to smoking and mode of delivery.  Obstet Gynecol. 2002;99785- 794PubMedGoogle ScholarCrossref
Greenland  SDrescher  K Maximum likelihood estimation of the attributable fraction from logistic models.  Biometrics. 1993;49865- 872PubMedGoogle ScholarCrossref
Hedley  AJPeters  JLam  TH  et al.  Air Pollution and Respiratory Health in Primary School Children in Hong Kong, 1989-1992: Report to the Environmental Protection Department, Hong Kong Government.  Hong Kong, China Dept of Community Medicine, University of Hong Kong1993;330
Lam  THChung  SFWong  CM  et al.  Youth Smoking: Knowledge, Attitudes, Smoking in Schools and Families, and Symptoms due to Passive Smoking: The Youth Smoking and Health Survey, 1994.  Hong Kong, China Hong Kong Council on Smoking and Health1995;
Lam  THHedley  AJChung  SF  et al.  Smoking, Passive Smoking and Respiratory Ill Health in Primary 3-6 Children: Smoking and Passive Smoking in Children, 1998.  Hong Kong, China Hong Kong Council on Smoking and Health1998;
Wakefield  MBanham  DMartin  JRuffin  RMcCaul  KBadcock  N Restrictions on smoking at home and urinary cotinine levels among children with asthma.  Am J Prev Med. 2000;19188- 192PubMedGoogle ScholarCrossref
Emerson  JWahlgren  DHovell  MMeltzer  SZakarian  JHofstetters  C Parent smoking and asthmatic children's exposure patterns.  Addict Behav. 1994;19677- 689PubMedGoogle ScholarCrossref
Blizzard  LPonsonby  ALDwyer  TVenn  ACochrane  JA Parental smoking and infant respiratory infection: how important is not smoking in the same room with the baby?  Am J Public Health. 2003;93482- 488PubMedGoogle ScholarCrossref
Hiley  CMorley  C Risk factors for sudden infant death syndrome: further change in 1992-3.  BMJ. 1996;3121397- 1398PubMedGoogle ScholarCrossref
Blackburn  CSpencer  NBonas  SCoe  CDolan  AMoy  R Effect of strategies to reduce exposure of infants to environmental tobacco smoke in the home: cross sectional survey.  BMJ. 2003;327257PubMedGoogle ScholarCrossref
Roseby  RWaters  EPolnay  ACampbell  RWebster  PSpencer  N Family and carer smoking control programmes for reducing children's exposure to environmental tobacco smoke [Cochrane review].  Cochrane Database Syst Rev. 2003;3CD001746PubMedGoogle Scholar
Vogt  TM Effects of parental smoking on medical care utilization by children.  Am J Public Health. 1984;7430- 34PubMedGoogle ScholarCrossref
McBride  CMLozano  PCurry  SJRosner  DGrothaus  LC Use of health services by children of smokers and nonsmokers in a health maintenance organization.  Am J Public Health. 1998;88897- 902PubMedGoogle ScholarCrossref
White  KLWilliams  TFGreenburg  BG The ecology of medical care.  N Engl J Med. 1961;265885- 892PubMedGoogle ScholarCrossref
Vogt  TMSchweitzer  SO Medical costs of cigarette smoking in a health maintenance organization.  Am J Epidemiol. 1985;1221060- 1066PubMedGoogle Scholar