screening for pregnancy risk-drinking

0 145-6OO8/94/1805- 1 156$3.00/0 ALCOHOLISM: CLINICAL AND EXPERIMENTAL RESEARCH

Vol. 18, No. 5 September/October 1994

Screening for Pregnancy Risk-Drinking Marcia Russell, Susan S. Martier, Robert J. Sokol, Pamela Mudar, Sidney Bottoms, Sandra Jacobson, and Joseph Jacobson

The efficacy of alcohol screening questionnaires, the TWEAK, T- ACE, NET, MAST, and CAGE, in detecting periconceptional risk- drinking, rl oz absolute alcohol/day, was investigated in 4743 African-American women attending an inner-city prenatal clinic who had reported ever drinking. Sensitivity, specificity, positive predictive value, efficiency, follow-up rates, and receiver operating characteristics of the questionnaires were examined to compare the overall effectiveness of the questionnaires and their performance at cut- points defining positive scores ranging from 1 to 3. Relatively little difference between TWEAK, 1-ACE, and MAST was seen in the receiver operating characteristic accuracy indices; NET and CAGE lagged behind. Sensitivity/specificity scores for the two questionnaires most sensitive at cut-point 1 were TWEAK (87/72) and 1-ACE (83/75). At cut-point 2, sensitivity was optimized with respect to specificity; TWEAK (79/83) was significantly more sensitive than T- ACE (70/85; p = 0.002). At cut-point 3, the two most sensitive tests were MAST (61/92) and TWEAK (59/94). In general, measures of merit were not greatly affected by the time between conception and the administration of the screens. Screening was most sensitive for women interviewed during the first 15 weeks of pregnancy; risk- drinkers tended to delay entry into prenatal care, increasing positive predictive values associated with screening later in pregnancy.

This study confirms the utility, when screening for risk-drinking during pregnancy, of brief questionnaires that assess alcohol intake indirectly by asking women about their tolerance to alcohol's eff8cts, psychological consequences of drinking, and significant others' con- cern about their drinking. It validates 1-ACE and provides preliminary data indicating that TWEAK may outperform 1-ACE.

Key Words: Alcohol Drinking, Pregnancy, Risk, Screening, African- American.

INCE DESCRIPTION of fetal alcohol syndrome in S 1973,' numerous studies have documented the potential of prenatal alcohol exposure to damage the fetus. The importance of intervening to reduce maternal drinking during pregnancy has been widely recognized.2 Because, fortunately, heavy drinking is relatively rare among pregnant women,3 it is necessary to screen large numbers to detect those whose alcohol use may pose a risk during pregnancy. Currently available biological markers of alcohol abuse are not only costly to assess, but they also lack sensitivity and ~pecificity.~ Accordingly, increased

From the Research Institute on Addictions (M.R., P.M.), Buffalo, New York; the Department of Obstetrics/Gynecology (S.S.M., S.B.), Wayne State UniversitylHutzel Hospital; and the Departments of Obstetrics1 Gynecology (R. J.S.) and Psychology (S.J.. J. J.), Wayne State University, Detroit, Michigan.

Received for publication December 13, 1993; accepted April 18, 1994 This study was supported by Grants ROl-AA07606 and ROl-AA06966

Reprint requests: Marcia Russell, Ph. D., Research Institute on Addic-

Copyright 0 I994 by The Research Society on Alcoholism.

from the National Institute on Alcohol Abuse and Alcoholism.

tions, 1021 Main Street, Bufalo, NY 14203.

1156

attention has been given to the use of brief questionnaires to screen for risk-drinking.

The two most widely used screening questionnaires, the Michigan Alcoholism Screening Test (MAST)' and the CAGE: were originally developed in male populations to detect alcoholism. Thus, there is a potential for differences in the sensitivity and specificity of these questionnaires related both to gender and the condition for which one is screening when they are applied in obstetric populations to screen for risk-drinking.' Concerns regarding the efi- cacy of available screening questionnaires led to the de- velopment of a new brief screening questionnaire based on research in obstetric populations, the T-ACE.' The unique contribution of T-ACE was its item on tolerance to alcohol's effects, designed to circumvent denial or underestimation often triggered by direct questions about alcohol intake.' Two brief questionnaires subsequently developed in obstetric populations, the TWEAK" and the NET," include the tolerance item as well.

The purpose of the present study is to compare the efficacy of the TWEAK, T-ACE, NET, MAST, and CAGE in detecting risk-drinking among obstetric patients. The comparison will take into consideration the influence on efficacy of cut-points used to define positive scores on the brief screening questionnaires and the time during pregnancy that screening is conducted.

METHODS

Sample Questionnaires were administered to women on their first visit to a

Detroit core city prenatal clinic by interviewers trained in eliciting alcohol history and consumption information. Analyses are based on consecutive patients who admitted having drunk alcohol at some time (n = 4743); all were African-American and of low socioeconomic status.

Measures MASF. The MAST consists of 25 questions, many used in previous

alcoholism surveys, and was developed to provide a quantitative, structured interview to screen for alcoholism that could be rapidly administered by professional as well as nonprofessional personnel. MAST items are weighted 0, 1,2, or 5; when summed, they yield scores ranging from 0 to 53.

CAGE. The CAGE consists of four items: Have you ever felt you ought to Cut down on your drinking? Have people Annoyed you by criticizing your drinking? Have you ever felt bad or Guilty about your drinking? Have you ever had a drink first thing in the morning to steady your nerves or get rid of a hangover (Eyeopener)? Each item receives a score of I for a positive response, for a possible total of 4 points.

T-ACE. T-ACE consists of four items, three of which it shares with Alcohol Clin Exp Res, Vol 18, No 5 , 1994: pp 1156-1 161

SCREENING FOR PREGNANCY RISK-DRINKING 1157

the CAGE (Annoy, Cut down, and Eyeopener).' In addition, it includes an item on Tolerance to the inebriating effects of alcohol, a question asking, "How many drinks does it take to make you feel high?" If women asked for clarification of the question, they were told, "This is how much it takes before you start to feel different than you usually do." Responses of three or more drinks were considered positive. Two points are scored for the Tolerance question, and 1 point is scored for each of the other three questions, for a possible total of 5 points.

TWEAK. The TWEAK consists of five items," three of which it shares with T-ACE (Tolerance, Eyeopener, and C/Kut down). The other two, shared with MAST, are: (1) Does your spouse (or parents) ever Worry or complain about your drinking? and (2) Have you ever awakened the morning after some drinking the night before and found that you could not remember a part of the evening before? (Amnesia or blackouts). Positive answers to the Tolerance and Worry questions score 2 points each, and the last three questions score 1 point each, for a possible total of 7 points.

NET. NET consists of three items, two of which it shares in common with T-ACE (Eyeopener and Tolerance)." In addition, it includes a question from the MAST Do you feel you are a Normal drinker? (That is, drink no more than the average.) The Tolerance question scores 2 points, and the other two questions score 1 point each, for a possible total of 4 points.

Periconceptional Risk-Drinking. Drinking during a typical week before women became pregnant was assessed using a procedure suggested by Bowman et al." It is comparable to a method currently known as the timeline follow-back ~r0cedure.l~ Beginning with a typical weekend (Friday PM), types of alcoholic beverages (beer, wine, wine coolers, brandy, or other liquor) and amount of each beverage drunk were recorded for each day of the week. Cognitive interviewing techniques were used to improve reporting. l4 For example, to help respondents recall their drinking on a typical Friday, interviewers were trained to ask them what they usually did on a Friday afternoon and evening (e.g., did they stay home, go out, have friends over, whatever), and whether they usually drank while they did this. Riskdrinking was defined as 1 oz or more of absolute alcohol/day (-14 or more drinks/week)?

Gestational Age at Screening. Time between conception and administration of the screens was assessed in terms of gestational age, which was based on the date of the last menstrual period. This information was available for all but 32 of the 4743 patients in the study.

Procedures All measures were assessed in all 4743 patients ( n = 471 1 for gesta-

tional age). Trained interviewers administered first the MAST and the CAGE, and then the timeline follow-back and the Tolerance question. This order was adopted in response to pilot studies that suggested that alcohol intakes were underreported when the timeline follow-back was administered before the MAST and the CAGE. The TWEAK, T-ACE, and NET were not administered as separate screening instruments; they were based on the Tolerance question and items embedded in the MAST and CAGE. It took -20 min to administer all the measures, most of which was devoted to the MAST and the timeline follow-back.

Analyses

The following measures of merit were estimated for each screen: sensitivity, the probability that a riskdrinker is positive on the test; spec$city, the probability that a non-riskdrinker is negative on the test; positive predictive value (PPV), the probability that a woman with a positive screening score is a risk-drinker, or the true-positive rate; and &cieecy, the overall percentage of women correctly identified with respect to ri~kdrinking.'~ The false-positive rate was calculated by sub- tracting PPV from 100. The follow-up rute (i.e., the number of positive screens that must be followed-up to identify risk-drinkers) was calculated by addingthe number of riskdrinkers positive on screening [true-positive rate X the number of riskdrinkers in the sample (n = 270)] to the

number of non-risk-drinkers positive on screening [false-positive rate x non-riskdrinkers in the sample (n = 4473)], and dividing by the total sample size ( n = 4743). The cut-point of a screening test is the value at or above which scores on the test are considered positive.

In addition, receiver operating characteristics (ROCs)16 for the screening questionnaires were investigated. For an ROC curve, the relationship between sensitivity and specificity is plotted over a test's range of possible cut-points; sensitivity is plotted on the y-axis, and 100 - specificity is plotted on the x-axis. An ideal screen would be 100% sensitive and 100% specific; on an ROC curve, this would correspond to the upper left comer of the graph. Accordingly, the cut-point at which an ROC curve comes closest to the upper left comer of the graph indicates the point at which sensitivity is optimized with respect to specificity. The area under an ROC curve provides an index of the accuracy of the screening test that can be used to compare the efficacy of screening questionnaires."

RESULTS

Maternal characteristics and substance use are summarized for risk-drinkers and non-risk-drinkers in Table 1. Risk-drinkers tended to be somewhat older and to have had more pregnancies and more births. On average, risk- drinkers were screened -3 weeks later in their pregnancies than non-risk-drinkers. Also, risk-drinkers were more likely than non-risk-drinkers to smoke, and they smoked more heavily.

ROC curves for all five questionnaires are plotted in Fig. 1. Curves for MAST and CAGE are offset to the right, so that individual curves and cut-points can be distin- guished. The general shape of all five ROC curves is similar; sensitivity increases rapidly as cut-points decrease, with relatively small decreases in specificity. Accuracy indices estimating the area under each curve are summarized in Table 2. The area under the ROC curve for the TWEAK test is larger than that for the other screens. However, the difference in area accounted for by TWEAK compared with T-ACE and MAST is not great; NET and CAGE performed less well. By inspection, the TWEAK at cut-point 2 is closest to the upper left comer of the graph, indicating the optimal combination of sensitivity and specificity. A cut-point of 2 has also been recommended for T-ACE.'

To facilitate comparison of the screens, measures of merit and follow-up rates are summarized for cut-points ranging from 1 to 3 (Table 3). At every cut-point, the five- item TWEAK was more sensitive than the four-item T-

Table 1. Maternal Characteristics by Periconceptional Risk-Drinking Status [Mean f SD (n = 4743)]

Riskdrinking Yes No

Maternal characteristics (n = 270) (n = 4473)

Maternal age (yr) Parity Gravidity Prepregnancy weight (Ibs) Gestational age (wk)' % smokers Cigaretteslday Absolute alcohol/day (02)

* n = 4711

~~~

27.9 f 5.6 1.8k 1.6 4.6 f 2.5

143.6 f 37.7 25.2 f 8.9

84.4 14.4 f 11.5 2.73 f 2.53

24.0 f 6.0 1.2k 1.4

142.8 f 37.2 22.4 f 8.9 47.2

6.2 f 9.2 0.12 f 0.20

3.2 f 2.1

1 I58 RUSSELL ET AL.

A

9

9 r

- TWEAK ........ Q ........ T-ACE ---&--. NET

o 10 20 30 40 im o 10 20 30 40 im 100-Specificity (In Percents)

Fig. 1. ROC curves: (A) TWEAK, T-ACE, and NET. (B) MAST and CAGE (n = 4743).

Table 2. Accuracv Indices for ROC Curves

Screening Accuracy index Questionnaire k E )

TWEAK

MAST NET CAGE

T-ACE 0.865 (0.014) 0.840 (0.015) 0.833 (0.01 6) 0.793 (0.017) 0.776 (0.013)

Table 3. Measures of Merit and Follow-Up Rates (in %) for Brief Screening Questionnaires, at Cut-Points from 1 to 3 (n = 4743)

Screening

(cut-points) Sensitivity Specificity PPV Efficiency rate questionnares Fdlow-up

TWEAK (1)

MAST (1) NET (1) CAGE (1)

TWEAK (2)

MAST (2) NET (2) CAGE (2)

MAST (3) TWEAK (3)

CAGE (3) NET (3)

T-ACE (1)

T-ACE (2)

T-ACE (3)

07 83 00 71 68

79 70 69 61 49

61 59 45 30 24

72 16 72 75 17 75 75 16 75 86 23 85 82 18 81

83 22 83 85 22 85 85 21 84 87 22 85 93 30 91

92 32 90 94 39 92 97 46 94 98 52 94 99 58 95

32 28 28 18 21

21 18 18 16 9

11 9 6 3 2

ACE, and T-ACE was more sensitive than the three-item NET. Given the sample size of 270 risk-drinkers, TWEAK was significantly more sensitive than T-ACE (79% compared with 70%; goodness-of-fit x2 = 9.74, p = 0.002) at the recommended cut-point of 2. Despite having 25 items, MAST was less sensitive to risk-drinking than TWEAK or T-ACE at cut-points of 1 or 2; however, at cut-point 3, MAST was somewhat more sensitive than TWEAK, but less specific. CAGE did not perform well at any cut-point.

To determine whether the efficacy of screening was influenced by the time between conception and the administration of the screens, measures of merit for the

TWEAK (using a cut-point of 2) were examined for women screened at different periods during pregnancy: <15 weeks, 15-24 weeks, 25-34 weeks, and 135 weeks. As summarized in Table 4, TWEAK was most sensitive to risk-drinking among patients screened before their 15th week of pregnancy, but after that there was little variability in sensitivity related to time of screening. Specificity, efficiency, and follow-up rates changed little over the course of pregnancy. However, PPVs increased with increasing gestational age. These increments were directly related to increases in the prevalence of risk-drinking associated with later entry into prenatal care and screening later in pregnancy.

DISCUSSION

This study validated the utility of T-ACE in screening for risk-drinking during pregnancy. A key finding was that TWEAK compared favorably to T-ACE and other brief questionnaires across a range of potential cut-points, com- bining high sensitivity levels with reasonable specificity levels. Screening later in pregnancy was associated with higher prevalence rates of risk-drinking and correspond- ingly higher PPVs. However, with the exception of women interviewed early in pregnancy, among whom screening seemed more sensitive, few other measures of merit were influenced by the time during pregnancy that the screens were administered.

The importance of the Tolerance item was confirmed by the fact that T-ACE performed almost as well as TWEAK and substantially better than CAGE, and the importance of items on feeling the need to K/Cut Down and others’ negative reactions to drinking was illustrated by the fact that TWEAK and T-ACE outperformed NET. The Worry item may be more sensitive than the Annoy item, because some women may not be annoyed if others worry or complain about their drinking.

Although MAST performed reasonably well in the present study, the most compelling argument against using MAST to screen for risk-drinking during pregnancy is that it is too long and difficult to score.8 Although the MAST is a useful research tool, screens must be short and easy to score if they are to be practical for clinical use. Indeed, structured interviews have been developed to assess DSM- 111-R diagnostic criteria for alcohol abuse or dependence that are shorter and no more difficult to score than the MAST.I8 Therefore, there would be no savings involved in using MAST to screen patients before administering a diagnostic interview. lo

It has been proposed in the past that women be asked directly about the quantity and frequency of their alcohol use.” Indeed, the fact that direct questions on alcohol intake were used in the present study to validate the brief screens would suggest the validity of such an approach. However, the timeline follow-back method used to assess alcohol intake in the present study is time-consuming and


TaMe 4. Measures of Merit and Follow-up Rates (in YO) for TWEAK (Cut-Point = 2) and Prevalence of Risk-Drinking by Gestational Age at the T im of Screening

Gestational FOIIOW-UP Prevalence of age (wk) no. Sensitivity Specificity PPV Efficiency rate riskdrinking

4 5 1204 91 04 10 04 19 3.7 15-24 1592 76 03 19 a2 20 5.2 25-34 1419 74 04 26 03 20 7.1 235 41 6 74 06 35 05 19 9.1

All 471 1 70 04 22 03 20 5.7

requires special training for its administration, making it more appropriate for research than clinical use, as is the case for the MAST. Also, as indicated earlier, pilot studies indicated that substantially lower rates of risk-drinking were reported when the timeline follow-back was administered before the screens. This indicates that the validity of alcohol consumption data is influenced by the context in which questions are asked and suggests that administration of screens asking about consequences of alcohol use that may have occurred in the past increases the validity of subsequent questions about intake. The importance of order effects on validity of screens was also documented in a study conducted at Walter Reed Army Medical Center to test the ability of CAGE to detect alcohol abuse or dependence.20 Sensitivity of the CAGE was reduced from 95% to 32% by asking quantity-frequency questions before administration.

Given recent efforts to alert women to the dangers of drinking during pregnancy, it seems reasonable to suppose that it is not socially desirable for pregnant women to admit large alcohol intakes. Indeed, this is the rationale for proposing that an oblique approach be taken by using brief screens to introduce the topic of alcohol consumption. This is not to say that direct questions on alcohol consumption are unimportant. Some risk-drinkers missed by brief screens, such as the TWEAK or T-ACE, will be picked up by asking directly about their drinking patterns. '' Not all obstetric patients minimize their drinking, and risk-drinking can sometimes be detected even in heavy drinkers who deliberately underreport, because an intake of 14 drinks/week may seem low to them.

Dramatic differences in the performance of all the questionnaires were associated with changes in their cut-points, illustrating the need to consider cut-point, as well as the questionnaire, when selecting a screening method. It is well known that lowering the cut-point increases the sensitivity of a screening method and decreases its specificity." Sensitivity is usually given priority when screening; therefore a case can be made for selecting either TWEAK (1) or T-ACE (l) , both of which had high sensitivity. However, when a condition is relatively rare, small decreases in the specificity can greatly increase the false- positive rate, and, consequently, increase the follow-up rate. Thus, for TWEAK (I), 32% of the obstetric population must be followed-up when sensitivity is 87% and specificity is 72% (false-positive rate = 84%), and for TWEAK (3), only 9% of the patients must be followed-

up if sensitivity is 59% and specificity is 94% (false-positive rate = 61%).

The cost of following-up patients who screen positive is thought to be offset by the benefits of reducing risk- drinking during pregnancy; however, if other prevention programs are competing for scarce resources, it may be necessary to justify carefully the level of effort devoted to this aspect of prenatal care. In the present study, with 270 risk-drinkers and 4473 non-risk-drinkers, every 1 Q o incre- ment in sensitivity meant that 2.7 more risk-drinkers were positive on a screening questionnaire; every 1% decrease in specificity meant that an additional 44.7 non-risk- drinkers were positive; and every increase of 1% in the follow-up rate meant that an additional 47.4 patients had to be followed-up. Thus, lowering the cut-point for TWEAK from 2 to 1 increased sensitivity by 8% (+22 risk-drinkers), but decreased specificity 1 1 % (+533 non- risk-drinkers). Similarly, lowering the cut-point of T-ACE from 2 to 1 identified an additional 36 risk-drinkers at the cost of following-up an additional 500 non-risk-drinkers.

It is our hope that these promising initial studies will stimulate additional research on the TWEAK, T-ACE, and NET. There are several questions that warrant further investigation. One is the utility of self-administered ver- sions of the brief questionnaires. In the present study, literacy levels were low, which made it necessary for interviewers to administer the screens. However, self-administered alcohol questionnaires have been used exten- sively in gynecologic2' and obstetric2' patients in Buffalo, NY, and a self-administered form of the TWEAK was used successfully in samples of the general household population and in hospital clinic outpatient^.^^ The merits of self- versus interviewer-administered brief questionnaires are well worth investigating in obstetric populations, because previous research suggests that women may report alcohol-related behavior more fully in self-administered questionnaires than when asked about it directly by their physician^.'^ Brief alcohol screening questionnaires could also be incorporated into interactive computer programs designed to inventory health-related beha~ior.'~

A critical question is whether measures of merit will be influenced by administering the TWEAK, T-ACE, and NET independently, rather than constructing them from items embedded in the MAST and CAGE. As previously mentioned, the MAST is a lengthy questionnaire that asks about prior treatment for alcoholism and social consequences, such as fights after drinking and arrests for driv- ing while intoxicated, that are more typical of male than

I160 RUSSELL ET AL.

female alcoholics. These MAST questions may either de- sensitize or sensitize obstetric patients. Denying a treatment history and social consequences of alcoholism may make women feel more comfortable about reporting psychological consequences of drinking, exemplified by the Worry, Cut Down, and Annoy items, or conversely, they may feel the need to distance themselves from such behavior by denying any alcohol-related consequences. In another group of obstetric patients (n = 1444) in which T-ACE was administered independently, sensitivity was lower than in the present study, 60% compared with 70%, suggesting that desensitization may take place when the questions are embedded. Specificity was 85% in both samples.

In the study of the self-administered form of the TWEAK mentioned herein,23 TWEAK was administered independently and performed well, but it followed administration of the Brief-MAST and the CAGE, and the possibility that this desensitized respondents cannot be ruled out. Perhaps it will prove optimal to combine desensitization procedures with the oblique approach to screening by asking first a few questions about past, then present, severe forms of substance abuse, such as illicit drug use or having been treated for alcohol problems; next, questions about current, less negative alcohol-related experiences, such as psychological consequences of drinking; next, questions on periconceptional drinking (e.g., drinking in the past, before pregnancy), as was done in the present study; and only then asking about current intakes.26

A minor point concerns the wording of two CAGE items, Eyeopener and K/Cut Down, which were used in constructing TWEAK, T-ACE, and NET. The CAGE version of these items includes the past (Have you ever. . . .), whereas it has been proposed elsewhere” that they be cast in the present (Do you sometimes. . .). Thus, some women may have been mistakenly identified as current risk-drinkers because of morning drinking or feeling the need to cut down in the past. Research is needed to determine whether specificity with respect to risk-drinking in the current pregnancy can be improved by wording these items in the present, without loss in sensitivity.

Finally, the present study is based on a disadvantaged, African-American population attending an inner-city prenatal clinic in Detroit, MI. It remains to be demonstrated whether comparable results would be obtained in other regions of the country in obstetric populations differing in their sociodemographic characteristics.

Despite the fact that it has been over 20 years since fetal alcohol syndrome was described in the U.S., many patients, particularly African-Americans, still report not being counseled about drinking during pregnancy.” The present work is part of an ongoing program to develop and document the value of screening procedures convinc- ingly enough so that they will be widely implemented. We have identified screens that have proven to be effective, brief, and easy to score. We have also identified factors

that may influence screening for risk-drinking during pregnancy and that are in need of further research.

REFERENCES 1. Jones KL, Smith D W Recognition of the fetal alcohol syndrome

in early infancy. Lancet 2:999-1001, 1973 2. US. Attorney General: Surgeon General’s advisory on alcohol and

pregnancy. FDA Drug Bull 1 1:MS3 18, 198 1 3. Abel EL, Sokol RJ: A revised conservative estimate of the incidence

of FAS and its economic impact. Alcohol Clin Exp Res 15:514-524, 1991

4. Chan AWK: Biochemical markers for alcoholism, in Miller NS (ed): Comprehensive Handbook of Drug and Alcohol Addiction. New York, Marcel Dekker, 1991, pp 31 1-338

5. Selzer M L The Michigan Alcoholism Screening Test: The quest for a new diagnostic instrument. Am J Psychiatry 127:1653-1658, 1971

6. Ewing JA: Detecting alcoholism: The CAGE questionnaire. J A W

7. Russell M, Chan AWK, Mudar P Gender and screening for alcohol problems, in Wilsnack R, Wilsnack S (eds): Gender and Alcohol. Pisca- taway, NJ, Rutgers University Press (in press), 1994

8. Sokol RJ, Martier SS, Ager JW: The T-ACE questions: Practical prenatal detection of risk-drinking. Am J Obstet Gynecol 160:863-870, 1989

9. Sokol RJ, Martier S , Ernhart C: Identification of alcohol abuse in the prenatal clinic, in Chang NC, Chao HM (eds): NIAAA Research Monograph 17: Early Identification of Alcohol Abuse (DHHS Publica- tion No. ADM 85-1258). Washington, D.C., U.S. Government Printing

10. Russell M New assessment tools for drinking in pregnancy: T- ACE, TWEAK, and others. Alcohol Health Res World (in press), 1994

1 1. Bottoms SF, Martier SS, Sokol RJ: Refinements in screening for risk drinking in reproductive-aged women: The “net” results. Alcohol Clin Exp Res 13:339, 1989 (abstr)

12. Bowman RS, Stein LI, Newton JR: Measurement and interpre- tation of drinking behavior. I. On measuring patterns of alcohol consumption. 11. Relationships between drinking behavior and social ad- justment in a sample of problem drinkers. J Stud Alcohol 36: 1154-1 172, 1975

13. Sobell LC, Sobell MB: Timeline follow-back A technique for assessing self-reported alcohol consumption, in Litten R, Allen J (eds): Measuring Alcohol Consumption. Totowa, NJ, Humana Press, 1992, pp

14. Friedenreich CM: Improving long-term recall in epidemiologic studies. Epidemiology 5: 1-4, 1994

15. Hennekens CH, Buring J E Screening, in Mayrent SL (ed): Epi- demiology in Medicine. Boston, MA, Little, Brown & Company, 1987,

16. Metz CE: Basic principles of ROC analysis. Semin Nucl Med

17. Hanley JA, McNeil BJ: The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 143:29-36, 1982

18. Robins LN, Helzer JE, Croughan J, Ratcliff KS: National Institute of Mental Health Diagnostic Interview Schedule: Its history, characteristics and validity. Arch Gen Psychiatry 38:381-389, 1981

19. Rosett HL, Weiner L, Edelin KC: Treatment experience with pregnant problem drinkers. JAMA 249:2029-2033, 1983

20. Steinweg DL, Worth H: Alcoholism: The keys to the CAGE. Am

2 1. Russell M, Coviello D Heavy drinking and regular psychoactive drug use among gynecological outpatients. Alcohol Clin Exp Res 1 2 : W 406, 1988

22. Russell M, Skinner JB: Early measures of maternal alcohol misuse as predictors of adverse. pregnancy outcomes. Alcohol Clin Exp Res

23. Chan AWK, Pristach EA, Welte JW, Russell M: Use of the

252: 1905-1 907,1984

Ofice, 1985, pp 209-227

4 1-72

pp 327-341

81283-298, 1978

J Med 94520-523, 1993

12~824-830, 1988


TWEAK test in screening for alcoholism/heavy drinking in three populations. Alcohol Clin Exp Res 17:1188-1192, 1993

24. Russell M, Bigler L Screening for alcohol-related problems in an outpatient obstetric-gynecologic clinic. Am J Obstet Gynecol 134:4-12, 1979

screening by computer in a health maintenance o r g a n k t i o n - b d prenatal care clinic. Am J Obstet Gynecol 165506-514, 1991

26. Sokol RJ, Ager JW, Martier SS: Methodological issues in obtain-

ing and managing substance abuse information from prenatal patients, in Kilbey MM, Asghar K (eds): NIDA Research Monograph 1 17: Meth- odologcal I~~~~ in Epidemiolo,$d, hevention, and Treatment Re- search on and Their Children. Rockville, MD,

27. Kogan MD, Kotelchuck M, Alexander G, Johnson WE: Racial disparities in reported prenatal care advice from health care providers. Am J Public Health 84232-88, 1994

25. Lapham SC, a n g MK, Skipper B Prenatal behavioral risk u.s. Government Printing ofice, l9g2> PP 80-97

screening for pregnancy risk-drinking

Documents