Patients’ characteristics and providers’ attitudes: predictors of screening pregnant women for illicit substance use

Child Abuse & Neglect 28 (2004) 209–223

Patients’ characteristics and providers’ attitudes: predictors of screening pregnant women for illicit substance use夽 Bonnie D. Kerker a,b,∗,1 , Sarah M. Horwitz b,c,1 , John M. Leventhal d a

New York City Department of Health and Mental Hygiene, 125 Worth St., Room 315, CN #6, New York, NY 10013, USA b Department of Epidemiology and Public Health, Yale University School of Medicine, New Haven, CT, USA c Department of Psychiatry, Case Western Reserve University, Cleveland, OH, USA d Department of Pediatrics, Yale University School of Medicine, New Haven, CT, USA Received 19 December 2002; received in revised form 22 July 2003; accepted 31 July 2003

Abstract Objective: This study’s aim was to determine how patients’ and providers’ characteristics affect hospital providers’ decisions to screen pregnant and postpartum women for illicit substances. Methods: A retrospective design was used. Participants included all low-income women (N = 1,100) who delivered at an urban teaching hospital over a 12-month period and the providers (N = 40) who provided prenatal and delivery care for these women. The medical records of the women were abstracted to obtain demographic, medical, social, and substance use information. Providers were interviewed to obtain data on their attitudes. The outcome was a dichotomous indication of whether patients were screened for illicit substances. Results: Multivariate logistic regression analyses found that women who were single (OR = 7.1), Black (OR = 1.9), received prenatal care at the prenatal clinic (OR = 5.6), saw fewer providers (1.1), or had a placental abruption (OR = 15.8), preterm labor (OR = 3.0), inadequate prenatal care (OR = 4.9), a history of involvement with Child Protective Services (OR = 3.9), a high social/Mental Health Risk Factor Score (OR = 1.4), a past or present history of illicit drug use (OR = 6.7), or a present history of tobacco use (OR = 1.7) were more likely to be screened than women without these characteristics. Women whose providers scored medium (OR = 2.5) or high (OR = 2.5) on the Professionalism Scale were more likely to be screened than women whose providers scored low on this scale.

夽

This research was supported by National Institute of Mental Health Grant 5T32-MH15783, the Heaney Family Trust, and a Woodrow Wilson Dissertation Grant in Women’s/Children’s Health. ∗ Corresponding author. 1 At the time this research was conducted, Dr. Kerker was a pre-doctoral fellow and Dr. Horwitz was an Associate Professor at the Department of Epidemiology and Public Health, Yale University School of Medicine. 0145-2134/$ – see front matter © 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.chiabu.2003.07.004

210

B.D. Kerker et al. / Child Abuse & Neglect 28 (2004) 209–223

Conclusions: Providers’ decisions to screen pregnant women for illicit substance use are influenced by both patients’ characteristics and providers’ personal attitudes. Hospital protocols might help reduce the potentially biased impact of attitudes on screening decisions. © 2004 Elsevier Ltd. All rights reserved. Keywords: Illicit drug screening; Decision-making; Provider attitudes; Substance use during pregnancy

Introduction Appropriate care of women and newborns requires accurate information about prenatal substance use. To identify cases of women using illicit substances during pregnancy, both pregnant women and those giving birth often are questioned by their medical providers. Substance use reporting by patients, however, has been found to be unreliable (Norton-Hawk, 1997; Ryan et al., 1994; Zuckerman et al., 1989). Consequently, urine toxicology screens of pregnant or delivering women may be employed to gauge the use of illicit substances in this population. Alternatively, newborns can be screened to detect prenatal substance use. Since newborns rarely exhibit many of the signs of exposure that would prompt screening, however, decisions to screen infants usually are based on either maternal factors or the results of maternal screens. In Connecticut, as in most states, there is no law mandating the universal screening of delivering mothers or their infants, and most hospitals have a selective screening process, based largely on providers’ clinical judgment. The consequences of providers’ decisions to screen extend far beyond medical outcomes and may include the involvement of the Child Protective Service (CPS) agency in the lives of women and children. Despite the importance of these decisions, previous research conducted to explain the factors that influence medical providers’ decisions to test women for illicit substance use is limited. Some studies have investigated testing decisions by examining the characteristics of patients that hospital providers report to influence their decisions. These factors, however, do not completely account for providers’ decisions, and most previous research has not included examinations of providers’ attitudes or perceptions of the organizational environment in which they work, each of which may influence screening decisions. Rather than rely on the factors that providers indicate are used to make screening decisions, this study examined the association between patient factors documented in the medical records of low-income women and their providers’ decisions to screen. Additionally, this research examined the roles that providers’ attitudes and perceptions of organizational factors play in decision-making by supplementing patients’ information with data from in-person interviews with the clinicians who provided care to them.

Methods Sample This study utilized two samples: low-income pregnant women and the healthcare providers who cared for them prenatally and at delivery. Since toxicology screens are not ordered often among privately insured patients, we chose to study a publicly insured patient population to enable us to examine the factors that influence screening decisions. The samples are described in Table 1.

B.D. Kerker et al. / Child Abuse & Neglect 28 (2004) 209–223

211

Table 1 Characteristics of study samples Sample

Criteria

How identified

Number eligible

Number (%) included

Collection method

Low-income women

Medicaid-eligible women who: (1) delivered at the hospital and (2) were discharged between March 1, 1998 and February 28, 1999

1,083

Providers at the prenatal clinic and one local neighborhood health center who provided prenatal and delivery care for low-income women from March 1, 1998-February 28, 1999

1,100 (98.5) (the medical records that were not included either could not be located or were involved in legal proceedings) Physicians: 29 (74.4)

Medical record abstraction

Healthcare providers

Cases were identified from the hospital management information system database based on insurance status, discharge date, and diagnosis Providers who signed or gave an order in the medical record of a woman in the study were included in the sample

Physicians: 39

CNMa : 5 Social workers: 5 Coordination of care providers: 3 Total: 52 a

30-minute interviews

CNMa : 5 (100) Social workers: 4 (80) Coordination of care providers: 2 (66.6) Total: 40 (76.9)

CNM: certified nurse midwife.

Instruments Low-income women. The form used to abstract data from the medical records is a modification of one used in an earlier study by Leventhal and colleagues (1997) and was designed to gather the factors recommended by the existing literature to be markers of possible drug use during pregnancy. For example, studies cite a history of substance use, no prenatal care, unexplained placental abruption, premature labor, and history of a sexually transmitted disease (STD) as being associated with screening decisions (Birchfield, Scully, & Handler, 1995; Kwong & Ryan, 1997). Further, social factors, such as homelessness and inadequate prenatal care, as well as demographic factors, such as race and age, have sometimes been associated with positive toxicology screens (Benderskey, Alessandri, & Gilbert, 1996; Chasnoff, Neuman, Thornton, & Callaghan, 2001; Frank et al., 1988; Ryan et al., 1994; Vega, Kolody, Hwang, & Noble, 1993). Although our study did not examine predictors of positive screening results, providers may base their screening decisions on factors related to positive screens, so we included these variables in our analyses. Consequently, the pre- and postnatal charts of eligible women were abstracted to obtain information on these demographic, medical, social, and self-reported substance use factors that may affect the decision to screen a woman. In addition, we hypothesized that although one social risk factor might not be predictive of screening, a combination of several, particularly in the presence of a mental health condition, might

212

B.D. Kerker et al. / Child Abuse & Neglect 28 (2004) 209–223

be a significant contributor. To account for this, a composite social/Mental Health Risk Factor Score was created (range 0–5), summing the number of the following five self-reported factors that were indicated in the medical records: family violence (e.g., spousal abuse), involvement of the criminal justice system, financial difficulty, inadequate housing (e.g., homelessness), and present mental health problems. Variables were recorded as “present history” if they were positive during this pregnancy and “past history” if they were positive prior to this pregnancy. With the exception of physical health conditions and CPS involvement, if no information was available on whether a particular factor was present or absent, the variable was recorded as “not documented.” Physical health conditions and CPS involvement were coded as absent unless documentation indicated their presence. Both the instrument and the study protocol were approved by the medical school’s Institutional Review Board (IRB) before the abstractions were implemented. The medical record abstraction also collected data on the dependent variable, maternal toxicology screen. This variable was dichotomous and was coded as positive if a maternal screen was documented (even if the result was negative), either prenatally or at delivery, in the mother’s medical record. Providers. The interview instrument used for providers was a compilation of original questions and previously validated, modified, and newly developed scales. All scales were designed with Likert Scale answer choices, ranging from 1 to 5. Both the interview instrument and the accompanying protocol were approved by the medical school’s IRB. As with the chart abstraction form, the interview was designed to gather information on the factors highlighted in the current literature, namely, the importance of attitudes and organizational environment. Attitudes are important to consider in decision-making because they are believed to influence physicians’ practice styles (Wolff, 1989), which have been shown to guide their use of medical resources (Wennberg & Gittelsohn, 1973). Although most people use their attitudes about groups to simplify the decision-making process, physicians, who work under time constraints and thus need to make quick judgments about complex and often life-threatening issues, may be particularly likely to do so (van Ryn & Burke, 2000). In fact, certain attitudes, such as feelings of pride and confidence in the profession, may help providers make difficult decisions with uncertain outcomes (such as screening a pregnant or postpartum woman) and convince others of the appropriateness of their choices (Conners, Dawson, Arkes, & Roach, 1990). Deciding whether or not to administer a toxicology screen to detect illicit substance use among pregnant and postpartum women also may be affected by providers’ attitudes due to the moral judgments that such use often elicits among nonusers. In fact, past studies have shown that many physicians hold strong and negative attitudes towards substance users (Chappel, Jordan, Treadway, & Miller, 1977; Najman, Klein, & Munro, 1982). As a result, our provider instrument included scales measuring providers’ attitudes towards the effect of prenatal substance use on an infant, their attitudes towards substance-using women and the confidence and pride they have in their professional practice. The culture and climate of the organizations in which providers work also may influence screening decisions. Organizational culture has been defined as the shared norms in an organization (Cooke & Rousseau, 1988; Glisson, 2000; Shortell, Rousseau, Gillies, Devers, & Simons, 1991; Shortell et al., 1994). Organizational climate is described as employees’ perceptions of the impact of the work environment on their personal well-being, including factors such as how oppressive or bureaucratic the agency seems to employees (Glisson, 2000). Studies indicate that such organizational factors play as important a role in hospitals as in nonservice business and industrial organizations (Glisson, 1981; Martin, Peters, & Glisson, 1998; Sovie,

B.D. Kerker et al. / Child Abuse & Neglect 28 (2004) 209–223

213

Table 2 Provider scales used in interviews Scale

Construct

Examplea

Citation

Questions

Alpha reliability coefficient

Risk marker

How indicative those interviewed believed drug use (and other nonsubstance related) behaviors to be of a harmful environment for a child Perception of the harmfulness of substance use to a fetus How irresponsible substance-using women are believed to be Perception of autonomy of the provider

How indicative do you think cocaine use during pregnancy is of a harmful environment for a child?

New

5

.80

How harmful do you think cocaine use during pregnancy is to a fetus? Anyone who uses drugs is irresponsible

New

2

.76

New

3

.89

Most of us are encouraged to use our own judgement.

5

.71

Organizational Climate

Oppressiveness of bureaucracy in the organization

Rules and regulations often get in the way of getting things done.

6

.72

Professionalism

Perception of confidence and professional pride

Other professions are actually more vital to society than mine.

Modified version of Glisson and James (2002) Modified version of Glisson and James (2002) Snizek (1972)

5

.78–.84

Harmfulness

Responsibility

Organizational culture

a

Providers were asked to use a 5-point scale to answer the question or respond to the statement.

1993). Consequently, the interview also included measures of culture and climate. Table 2 describes the scales used in the interview to measure providers’ attitudes and perceptions of organizational environment. In addition, the interviews collected data on providers’ personal factors, including their demographic characteristics, time working in the field, positions/job titles, time working at the current job, religious affiliation, race, and age. Data analysis All statistical analyses were conducted using SAS, version 6.12. Following an examination of inter-item correlations, all previously validated scales were constructed according to the instructions supplied by the authors, and newly created scales were constructed by adding the individual items together to reach a total score for each scale. Next, a factor analysis was conducted among the newly created scales’ questions to insure that the questions proposed as scales were, indeed, related to each other. Prior to hypothesis testing, univariate analyses (means, standard deviations, and frequencies) were calculated for each scale. Then, the medical record and provider interview databases were combined. For each provider documented in the medical record as having seen a patient, the corresponding provider’s personal factors,

214

B.D. Kerker et al. / Child Abuse & Neglect 28 (2004) 209–223

attitudes, and perceptions of the organizational environment were imported from the provider database. Since screening decisions often are made jointly by many providers and since the providers who ordered the screens were not necessarily those who made the screening decisions, the individual scales described above were averaged among all the providers who saw a patient either prenatally or at delivery. Information on a provider who saw a patient more than one time was counted only once. Although by using this methodology one outlier score could disproportionately influence the averaged score, no provider could be identified as the main decision-maker or as one who had more influence on the decision than the others. To adjust for the fact that not all patients saw the same number of providers, a variable that counted the number of providers seen by each patient was created and included in the analyses. Similarly, to account for contextual differences that may exist between the prenatal sites, indicator variables for each place of prenatal care (a yes/no variable for each site) were created for each patient and included in the analyses as well. As a result of this methodology, each observation (woman) in the joint database had a composite score for each provider scale and construct, reflecting the providers seen by that patient. These composite scores were then split into terciles (high score, medium score, low score) and categorized into indicator variables (yes/no for each tercile) for analytic purposes. Regarding providers’ categorical personal factors (gender, race/ethnicity, professional position), variables were created to reflect the proportion of a woman’s providers that fell into each category. Bivariate analyses were examined between the outcome variable (toxicology screens) and each independent variable. Since all of the variables being analyzed were categorical, χ2 tests were conducted. Next, a multivariate logistic regression analysis was conducted with the joint database, which included information on 1,083 women and their providers. The outcome variable was a dichotomous indication of whether women were screened for illicit drug use either during pregnancy or at labor and delivery. For all the categorical independent variables, indicator variables (yes/no) were created for each of the possible response categories. The logistic regression model to determine predictors of screening was built in a forward stepwise manner. Independent variables were added to the model in order of importance, based on the significance of the variable in the literature. Variables whose bivariate analyses were not statistically significant were not included in the modeling process. First, maternal factors were entered, followed by providers’ attitudes and perceptions of the organizational environment. Independent variables that did not contribute significantly to the model (at the alpha = .05 level) were individually deleted. Then, variables (such as providers’ personal characteristics, patient’s marital status, and patient’s prior CPS history) that the literature indicated might be associated with the independent or dependent variables were evaluated for confounding influences. Variables were considered confounders if they changed the parameter estimate of an independent variable by more than 10% and were statistically associated with both the outcome and the affected independent variable. As mentioned above, the models also included variables indicating the number of providers seen and the place of prenatal care.

Results Patient characteristics As Tables 3 and 4 indicate, most women were single, over 18 years of age, had less than a high school education and were unemployed; there were more Black women than White or Hispanic women. Prenatal

Table 3 Description of demographic and medical conditions of women (N = 1,083)a N (%) screenedc

N (%) not screenedc

χ2 (p value)

–d

108 (10.0)

4 (2.0) 129 (17.4) 1 (5.9) 3 (21.4) 2 (50.0)

195 (98.0) 612 (82.6) 16 (94.1) 11 (78.6) 2 (50.0)

36.18 (.001)

451 (41.6) 307 (28.3) 249 (23.0) 16 (1.5) 11 (1.0)

–d

4.9 (4.5)

88 (19.5) 26 (8.5) 32 (12.9) 0 (.0) 2 (18.2)

363 (80.5) 281 (91.5) 217 (87.2) 16 (100.0) 9 (81.8)

21.73 (.001)

Age ≤18 19–29 30+

94 (8.7) 758 (70.1) 230 (21.3)

–d

1 (.1)

14 (14.9) 94 (12.4) 43 (18.7)

80 (85.1) 664 (87.6) 187 (81.3)

5.90 (.052)

Place of prenatal caree Hospital prenatal clinic Neighborhood health center Other hospital clinic Other

660 (60.9) 126 (11.6) 7 (.6) 171 (15.8)

–d

119 (11.0)

108 (16.4) 16 (12.7) 1 (14.3) 5 (2.9)

552 (83.6) 110 (87.3) 6 (85.7) 166 (97.1)

21.11 (.001)

Present N (%)

Marital status Married Single Divorced Separated Widowed

199 (18.4) 741 (68.4) 17 (1.6) 14 (1.3) 4 (.4)

Race Black White Hispanic Asian Other

Placental abruption Preterm laborf Sexually transmitted disease

20 (1.8) 157 (14.5) 240 (22.2)

Absent N (%)

1,063 (98.2)

–d

10 (50.0)

10 (50.0)

22.08 (.001)

926 (85.5)

–d

42 (26.8)

115 (73.3)

25.11 (.001)

843 (77.8)

d

47 (19.6)

193 (80.4)

8.18 (.004)

–

B.D. Kerker et al. / Child Abuse & Neglect 28 (2004) 209–223

Not documented N (%)b

Variable

a

Due to rounding error, percentages in each category do not always add to 100%. “Not documented” indicates that the information was not documented in the medical record. c Of those for whom information was documented. d Not a coding option. e Analysis limited to those with documented prenatal care. f Delivery at <37 weeks gestation. b

215

216

Table 4 Description of social and substance abuse factors of women (N = 1,083)a Present N (%)

Inadequate prenatal cared

120 (11.1)

Past Child Protective Service involvement Social/Mental Health Risk Factor Score 0 1 2 3 4

55 (5.1) 757 (69.9) 218 (20.1) 76 (7.0) 22 (2.0) 10 (.9)

Absent N (%)

Not documented N (%)b

N (%) screenedc

N (%) not screenedc

χ2 (p value)

963 (88.9)

–e

44 (36.7)

76 (63.3)

58.08 (.001)

1,028 (94.9)

–

e

35 (63.6)

20 (36.4)

119.25 (.001)

–e

–e

76 (10.0) 42 (19.3) 17 (22.4) 7 (31.8) 9 (90.0)

681 (90.0) 176 (80.7) 59 (77.6) 15 (68.2) 1 (10.0)

73.33 (.001)

Present cocaine usef

45 (4.2)

700 (64.6)

338 (31.2)

37 (82.2)

8 (17.8)

138.08 (.001)

Present other substance use

82 (7.6)

813 (75.1)

188 (17.4)

41 (50.0)

41 (50.0)

95.07 (.001)

Present alcohol use

102 (9.4)

856 (79.0)

125 (11.5)

35 (34.3)

67 (65.7)

35.50 (.001)

Present tobacco use

366 (33.8)

608 (56.1)

109 (10.1)

93 (25.4)

273 (74.6)

60.93 (.001)

a

Due to rounding error, percentages in each category do not always add to 100%. “Not documented” indicates that the information was not documented in the medical record. c Of those for whom information was documented. d Less than three visits or late entry into care; includes patients with no documented care. e Not a coding option for these variables. f All substance use analyses include substance use known before screen, when applicable. b

B.D. Kerker et al. / Child Abuse & Neglect 28 (2004) 209–223

Variable

B.D. Kerker et al. / Child Abuse & Neglect 28 (2004) 209–223

217

care was obtained mostly at the prenatal clinic and the neighborhood health center, and the majority of women saw between 5 and 9 providers, with a mean number of 7.6 (range = 1–21). Over 10% of the women received inadequate prenatal care (less than three visits or late entry into care), and very few had some type of previous CPS involvement. Only 4.2% of the women had a known history of cocaine use during the index pregnancy, almost 8% were documented as using illicit substances other than cocaine during pregnancy, 9.4% of the women reported using alcohol during pregnancy, and more than one-third reported using tobacco during pregnancy. Provider characteristics Of the interviewed providers, 65.9% were female; 71% identified themselves as being White, nonHispanic, and 9.1% as African American. Overall, 68.2% of the interviewed providers were physicians, approximately 9% were social workers, 18.2% were certified nurse midwives, and 4.5% were care coordinators. The number of cases that providers saw during the study period ranged from 14 to 242 (mean = 107), and half the providers saw more than 93 cases in the study year. A description of the provider scales is provided in Table 5. Dependent variable A toxicology screen was performed on 151 cases (13.9%); 82 screens were done prenatally and 111 were done at delivery. Of these, 42 women were screened both prenatally and at delivery. Bivariate and multivariate analyses The bivariate analyses examining the relationships between screening decisions and both maternal factors and provider characteristics also are shown in Tables 3–4. These tables compare the percentage of women with a particular characteristic who were screened with the percentage of those with the characteristic who were not screened. When the multivariate logistic regression model was built, women who were Black (OR = 1.90) or had a placental abruption (OR = 15.79), preterm labor (OR = 3.02), inadequate prenatal care (OR = 4.88), a high social/Mental Health Risk Factor Score (OR = 1.36), a past or present history of illicit drug use (OR = 6.72), or a present history of tobacco use (OR = 1.71) were more likely to be screened than Table 5 Description of provider scalesa Variable

Mean

Standard deviation

Range

Risk marker Harmfulness Difficulty Responsibility Professionalism Climate Culture

16.10 8.44 13.68 8.31 11.02 19.42 12.90

2.91 1.54 1.47 3.59 2.87 4.31 3.77

5–20 3–10 10–15 3–15 6–18 10–27 5–19

a

The N for each variable varies with availability of data.

218

B.D. Kerker et al. / Child Abuse & Neglect 28 (2004) 209–223

Table 6 Final multivariate model predicting toxicology screens in joint databasea Variable

95% Confidence intervals

p value

7.12 1.90

2.25, 22.60 1.04, 3.47

.0009 .0378

5.59 1.11 15.79 3.02

2.77, 11.29 1.02, 1.21 4.32, 57.69 1.70, 5.36

.0001 .0162 .0001 .0002

Social factors Inadequate prenatal care Past involvement with CPSd Social/Mental Health Risk Factor Score

4.88 3.92 1.36

2.70, 8.81 1.69, 9.11 1.05, 1.78

.0001 .0015 .0194

Substance use Past or present illicit substance usee Present tobacco use

6.72 1.71

3.62, 12.48 3.62, 12.48

.0001 .0424

Providers’ attitudes Professionalism—mediumf Professionalism—highf

2.45 2.48

1.42, 4.22 1.33, 4.63

.0222 .0013

Demographic factors Singleb Blackc Medical factors Prenatal clinic patient Fewer providers seen Placental abruption Preterm labor

OR

a

Controlling for missing values and other categories of marital status and race/ethnicity Compared to married. c Compared to White. d Child Protective Services. e Includes substance use known before screen, when applicable. f Compared with low professionalism. b

women without these characteristics. Place of prenatal care, number of providers, marital status, and past history of CPS involvement were included as confounders, but also were significant predictors. Women who received prenatal care at the prenatal clinic (OR = 5.59), saw fewer providers (1.11), were single (OR = 7.12), and had a past history of CPS involvement (OR = 3.92) were more likely to be screened than women without these characteristics. Among the individual provider attitude scales, the Professionalism measure was the strongest statistical predictors of the outcome. Women whose providers scored mid- (OR = 2.45) or high level (OR = 2.48) on the Professionalism Scale were more likely to be screened than women whose providers scored the lowest on this scale (Table 6). The variables measuring organizational environment were not significant predictors in this model. Comment Discussion In this large retrospective sample, there were demographic (single marital status, Black race), medical (attendance at the prenatal clinic, attention of few providers, placental abruption, preterm labor), social

B.D. Kerker et al. / Child Abuse & Neglect 28 (2004) 209–223

219

(inadequate prenatal care, past history of CPS involvement, high composite social/Mental Health Risk Factor Score), and substance use-related (present history of illicit substance use, present history of tobacco use) factors associated with an increased likelihood of being given a toxicology screen. Although some factors that we found to be associated with providers’ screening decisions have been cited as screening criteria in prior research, other factors were different from those that have been discussed previously as being significant predictors. This discordance may be because of differences in the populations studied. Alternatively, it could be due to differences in the methodology used in each study. While most previous research has relied on provider recall, this study’s strength was its use of data from medical records and interviews with providers. Previous literature has, for example, cited placental abruption, preterm labor, inadequate prenatal care, and known substance use as influencing screening decisions (Birchfield et al., 1995; Kwong & Ryan, 1997). We also found these variables to be predictive in our sample. In contrast, although previous studies have highlighted the importance of a history of STDs as a predictor of screening, we did not find such a relationship in our multivariate analyses. Several variables that generally have not been indicated as predictors of screening decisions in the past were found to predict screening in our study. For example, while race has not been mentioned in most previous research examining screening decisions, in our data, which controlled for income, Black women were more likely to be screened than White women. Although some past studies have found significant associations between substance use and Black race, others have not indicated such a relationship (Chasnoff, Landress, & Barrett, 1990; National Institute on Drug Abuse, 2001; Richardson & Day, 1991; Ryan et al., 1994; Vega et al., 1993). Our results may represent providers’ accurate detection of an association between substance use and race. Or, the findings may be due to providers’ inaccurate perceptions of the Black population, indirectly suggesting that providers’ attitudes influence their decisions to screen women. Similarly, the positive relationship between a high composite social/Mental Health Risk Factor Score and toxicology screens may be indicative of an association among social factors, mental health conditions, and positive screens. Alternatively, this finding may suggest that providers’ attitudes about certain types of social and mental health characteristics affect screening decisions. While previous research has not directly examined the effects of providers’ attitudes on screening decisions, our results explicitly indicated that certain provider attitudes influence such decision-making. In these data, women whose providers felt high levels of professionalism were more likely to be screened than women whose providers felt lower levels. The Professionalism Scale asked about feelings of professional pride and confidence. This research, then, suggests that providers who expressed high levels of pride and confidence in their profession’s role are more likely to screen than those who expressed lower levels of these characteristics. Neither of the variables explicitly describing organizational characteristics was significantly associated with screening decisions in our study. While it may be that no relationship exists among these factors, the scales used to measure culture and climate might not have been sensitive enough to detect an association in this population. This might be because modified versions of the scales, not the full scales as they were developed, were used. In fact, Glisson and James (2002) found both original scales to significantly predict behavior among child welfare and juvenile justice case managers. Alternatively, there might not have been enough variation among hospital providers on these factors to detect associations in these data. Interestingly, although place of prenatal care was included as a confounder, it was a significant predictor of screening. Women who received prenatal care at the hospital’s prenatal clinic were more likely to be screened than women who sought care elsewhere. While the significance of this variable does not provide

220

B.D. Kerker et al. / Child Abuse & Neglect 28 (2004) 209–223

evidence that providers’ perceptions of the organizational environments in which they work influence their screening decisions, this finding may be an indirect indication that this is the case. Or, there may be differences among providers who choose to work at the different sites, which could affect the prevalence of screening at each site. Limitations Although this research highlighted some important features of screening decisions among providers, it had certain limitations. First, this was a retrospective study that relied on information recorded in medical records. Our findings, therefore, are only as good as the information provided by patients and providers in the medical records. Many of the patient factors of interest (particularly regarding substance use and social issues), in fact, were not well documented in the medical records. Since there is no standard mechanism for collecting these data at this hospital, biases in deciding to collect data could be related to screening decisions. Second, although the response rate for the hospital providers’ interviews was high (76%), the provider data set was quite small. In addition, it only represented a portion of the providers seen by the women, since private providers and those from other prenatal care sites were not included. As a result, the provider-based analyses may not have identified all existing relationships, and studies with a larger sample of providers should be conducted so that the factors associated with providers’ screening practices can be reported more confidently. Third, since patients saw many providers and we could not identify the key providers who cared for each patient, the attitude scores were composites over many individuals. This averaging may have masked some important relationships. Future researchers should develop criteria to determine the key providers in patients’ care, so that the characteristics of fewer providers per case can be evaluated in joint analyses. Fourth, this study only examined low-income women, so our results can only be generalized to this population. Since there may be a bias towards screening this group, it is important to understand differences between factors that predict screening in public and private patients. Future research should be conducted with samples large enough to predict screening among both private and public patients. Summary and implications Despite these limitations, we conclude that providers’ decisions to screen pregnant women for illicit substance use are influenced by both their patients’ characteristics and providers’ personal attitudes. Many of the patients’ factors found here to predict providers’ screening decisions have been cited in previous research. Other characteristics, perhaps due to differences in methodology, represent new and significant additions to the literature. These factors—patients’ race, social/Mental Health Risk Factor Score, and place of prenatal care—may be indirect indications of the impact that providers’ attitudes have on their decision-making. Further, while most previous research has focused on using patients’ factors alone to describe screening decisions, the current analysis indicates that direct measures of providers’ attitudes (feelings of professionalism) are important when considering clinical decision-making. Given the potential consequences of screening pregnant and postpartum women for illicit substance use, it is important not to screen women based on providers’ personal attitudes. One possible way to avoid that would be to institute a universal screening policy. There is much debate, however, over whether such a policy should be implemented (Birchfield et al., 1995). Universal testing, for example, is expensive and is often seen as a questionable, and perhaps unethical, use of scarce resources (Gustavsson, 1992).

B.D. Kerker et al. / Child Abuse & Neglect 28 (2004) 209–223

221

Within a selective screening policy, it is important that screening protocols be implemented. Without clear guidelines, it is understandable that personal feelings may affect screening decisions. The influence of such factors, however, can be discriminatory. Screening protocols that incorporate the factors clearly identified in both the literature and local data as being associated with positive toxicology screens might help avoid biased decisions. Our results emphasize the importance of using available protocols, and developing new ones when necessary, in all hospitals that employ selective screening policies. Acknowledgments The authors gratefully acknowledge the substantive contributions that Dr. Mark Schlesinger and Dr. Brian Forsyth made to this research, as well as earlier versions of this paper. References Bendersky, M., Alessandri, S., Gilbert, P., & Lewis, M. (1996). Characteristics of pregnant substance abusers in two cities in the northeast. American Journal of Drug and Alcohol Abuse, 22(3), 349–362. Birchfield, M., Scully, J., & Handler, A. (1995). Perinatal screening for illicit drugs: Policies in hospitals in a large metropolitan area. Journal of Perinatology, 15, 208–214. Chappel, J. N., Jordan, R. D., Treadway, B. J., & Miller, P. R. (1977). Substance abuse attitude changes in medical students. American Journal of Psychiatry, 134(4), 379–384. Chasnoff, I. J., Landress, H. J., & Barrett, M. E. (1990). The prevalence of illicit-drug or alcohol use during pregnancy and discrepancies in mandatory reporting in Pinellas County, Florida. The New England Journal of Medicine, 322, 1202–1206. Chasnoff, I. J., Neuman, K., Thornton, C., & Callaghan, M. A. (2001). Screening for substance use in pregnancy: A practical approach for the primary care physician. American Journal of Obstetrics and Gynecology, 184(4), 752–758. Conners, A. F., Dawson, N. V., Arkes, H. R., & Roach, M. J. (1990). Decision making in support: Physician perceptions and preferences. Journal of Clinical Epidemiology, 43(Suppl.), 59S–62S. Cooke, R. A., & Rousseau, D. M. (1988). Behavioral norms and expectations. Group and Organizational Studies, 13, 245–273. Frank, D. A., Zuckerman, B. S., Amaro, H., Aboagye, K., Bauchner, H., & Cabral, H. (1988). Cocaine use during pregnancy: Prevalence and correlates. Pediatrics, 8, 888–895. Glisson, C. (1981). A contingency model of social welfare administration. Administration in Social Work, 5(1), 15–29. Glisson, C. (2000). Organizational climate and culture. In Patti, R. (Ed.), The handbook of social welfare administration. Thousand Oaks, CA: Sage Publications. Glisson, C., & James, L. R. (2002). The cross-level effects of culture and climate in human service teams. Journal of Organizational Behavior, 23(6), 767–794. Gustavsson, N. S. (1992). Drug exposed infants and their mothers: Facts, myths and needs. Social Work in Health Care, 16, 87–101. Kwong, T. C., & Ryan, R. M. (1997). Detection of intrauterine illicit drug exposure by newborn drug testing. Clinical Chemistry, 43, 235–242. Leventhal, J. M., Forsyth, B. W. C., Qi, K., Johnson, L., Schroeder, D., & Votto, N. (1997). Maltreatment of children born to women who used cocaine during pregnancy: A population based study. Pediatrics, 100(2), E7. Martin L. M., Peters, C. L., & Glisson, C. (1998, December). Factors affecting case management recommendations for children entering state custody. Social Service Review, 72(4) 521–543. Najman, J. M., Klein, D., & Munro, C. (1982). Patient characteristics negatively stereotyped by doctors. Social Science an Medicine, 16, 1781–1789. National Institute on Drug Abuse. (2001). Pregnancy and drug use trends. Available: http://www.nida.nih.gov/infofax/ pregnancytrends.html Norton-Hawk, M. A. (1997). Frequency of prenatal drug abuse: Assessment, obstacles and policy implications. Journal of Drug Issues, 27, 447–462.

222

B.D. Kerker et al. / Child Abuse & Neglect 28 (2004) 209–223

Richardson, G. A., & Day, N. L. (1991). Maternal and neonatal effects of moderate cocaine use during pregnancy. Neurotoxicology Teratology, 13, 455–460. Ryan, R. M., Wagner, C. L., Schultz, J. M., Varley, J., DiPreta, J., & Sherer, D. M. (1994). Meconium analysis for improved identification of infants exposed to cocaine in utero. Journal of Pediatrics, 125, 435–440. Shortell, S. M., Rousseau, D. M., Gillies, R. R., Devers, K. J., & Simons, T. L. (1991). Organizational assessment in intensive care units (ICUs): Construct development, reliability and validity of the ICU nurse-physician questionnaire. Medical Care, 29, 709–726. Shortell, S. M., Zimmerman, J. E., Rousseau, D. M., Gillies, R. R., Wagner, D. P., & Draper, E. A. (1994). The performance of intensive care units: Does good management make a difference? Medical Care, 32, 508–525. Snizek, W. E. (1972). Hall’s Professionalism Scale: An empirical reassessment. American Sociological Review, 37, 109–114. Sovie, M. D. (1993). Hospital culture—Why create one? Nursing Economics, 11, 69–90. van Ryn, M., & Burke, J. (2000). The effect of patient race and socio-economic status on physicians’ perceptions of patients. Social Science and Medicine, 50, 813–828. Vega, W., Kolody, B., Hwang, J., & Noble, A. (1993). Prevalence and magnitude of perinatal substance exposure in California. The New England Journal of Medicine, 329, 850–854. Wennberg, J., & Gittelsohn, A. (1973). Small area variations in health care delivery. Science, 182, 1102–1108. Wolff, N. (1989). Professional uncertainty and physician medical decision-making in a multiple treatment framework. Social Science & Medicine, 28(2), 99–107. Zuckerman, B., Frank, D. A., Hingson, R., Amaro, H., Levenson, S. M., & Kayne, H. (1989). Effects of maternal marijuana and cocaine use on fetal growth. The New England Journal of Medicine, 320(12), 762–768.

Résumé Objectif: Le but de cette étude consistait à déterminer dans quelle mesure les caractéristiques des patientes et des fournisseurs de soins pourraient affecter la décision de l’hˆopital de dépister chez les femmes enceintes ou ayant accouché la présence de substances illicites. Méthode: On a procédé sur un mode rétrospectif. Les participantes étaient toutes les femmes à bas revenu (N = 1,100) ayant accouché dans un hˆopital universitaire pendant une période de 12 mois ainsi que les fournisseurs de soins (N = 40) qui s’étaient occupés de ces femmes avant et après leur accouchement. On a résumé leurs dossiers médicaux afin d’obtenir des informations démographiques, médicales, sociales et concernant l’usage de stupéfiants. On a interviewé les fournisseurs de soin afin d’obtenir des données sur leurs attitudes. Le résultat a été une indication dichotomique concernant le dépistage de ces patientes selon l’usage de substances illicites. Résultats: Grˆace à des analyses de régression logistique à plusieurs variables, on a trouvé que les femmes qui étaient le plus susceptibles de faire l’objet d’un dépistage se trouvaient dans les cas suivants: seules (OR = 7.1), noires (OR = 1.9), suivies pendant leur grossesse à la clinique prénatale (OR = 5.6), avaient rencontré le moins de soignants (OR = 1.1), avaient présenté une rupture du placenta (OR = 15.8), ou un travail avant terme (OR = 3.0), avaient été mal suivies pendant leur grossesse (OR = 4.9), avaient été en contact avec les services de Protection de l’Enfance (OR = 3.9), ou encore présentaient un score de risque élevé au point de vue social et de la santé mentale (OR = 1.7). Les femmes à qui les soignants avaient donné des scores moyens (OR = 2.5) ou élevés (OR = 2.5) sur l’échelle de professionnalisme étaient plus susceptibles d’ˆetre dépistées que celles dont les scores étaient bas à cette échelle. Conclusion: La décision des fournisseurs de soins de soumettre les femmes à un dépistage au sujet de la consommation de substances illicites est influencée par les attitudes des patientes et des soignants. Utiliser des protocoles à l’hˆopital peut réduire le biais exercé par l’impact des attitudes sur la décision d’un dépistage.

B.D. Kerker et al. / Child Abuse & Neglect 28 (2004) 209–223

223

Resumen Objetivo: El objetivo de este estudio fue determinar como afectan las caracter´ısticas de los pacientes y de los profesionales en las decisiones sobre la evaluación de uso de sustancias il´ıcitas en mujeres embarazadas y en fase post-parto. Método: Se utilizó un diseño retrospectivo. Las participantes fueron todas las mujeres de bajos ingresos (N = 1,100) que dieron a luz en un hospital universitario urbano durante un periodo de 12 meses y los profesionales (N = 40) que atendieron a dichas mujeres en el parto y en la fase prenatal. Los registros médicos de las mujeres fueron resumidos para obtener datos médicos, sociales e información sobre uso de sustancias tóxicas. Los profesionales fueron entrevistados para obtener datos sobre sus actitudes. La variable dependiente fue la indicación dicotómica de si los pacientes recibieron o no una evaluación sobre el uso de sustancias il´ıcitas. Resultados: Los análisis de regresión log´ıstica multivariada señalan que las mujeres solteras (OR = 7.1), de raza negra (OR = 1.9), que recibieron atención en una cl´ınica prenatal (OR = 5.6), que vieron menos profesionales (1.1), tuvieron un parto prematuro (OR = 3.0), un inadecuado cuidado prenatal (OR = 4.9), una historia de implicación con los Servicios de Protección Infantil (OR = 3.9), o un factor de alto riesgo de tipo social o de salud mental (OR = 1.7) ten´ıan más posibilidades de ser evaluadas que las que no ten´ıan estas caracter´ısticas. Las mujeres cuyos profesionales obtuvieron una puntuación media (OR = 2.5) o elevada (OR = 2.5) en la Escala de Profesionalismo tuvieron más posibilidades de ser evaluadas que las mujeres cuyos profesionales puntuaron bajo en dicha escala. Conclusión: La decisión de los profesionales de evaluar a las madres sobre el uso de sustancias il´ıcitas está influidas tanto por caracter´ısticas de los pacientes como por las actitudes de los profesionales. Los protocolos hospitalarios pueden ayudar a reducir el impacto potencialmente sesgado de las actitudes en las decisiones sobre la realización de estas evaluaciones.