This is a critical abstract of an economic evaluation that meets the criteria for inclusion on NHS EED. Each abstract contains a brief summary of the methods, the results and conclusions followed by a detailed critical assessment on the reliability of the study and the conclusions drawn.

The health technology studied was an on-site support group for drug-abusing pregnant women.

Treatment.

Cost-effectiveness analysis.

The study population included pregnant women who abused substances such as drugs and alcohol.

The setting was an urban, academic hospital obstetric clinic. The study was conducted in the USA.

The effectiveness, resource use and charges data were gathered from 1989 to 1990.

The effectiveness data were derived from a single study.

The data on Medicaid charges were analysed for the same sample.

The study sample comprised 121 pregnant women who were registered for prenatal care before the 28th week of pregnancy, and had labour or delivery records available for chart review. Of these women, 54 were in the intervention group and 67 in the control group. All the women screened positive for alcohol and/or drug use based on clinician interview or urinalysis drug toxicology at the first prenatal visit. No power calculations were performed to determine the sample size. The baseline characteristics were given in order to compare how the study sample represented the study population.

A quasi-experimental study design was employed. The women followed the usual prenatal care and were assessed for their need of substance-abuse treatment by CAGE (mnemonic) screening questions and the Addiction Severity Index (ASI)). The women were then referred to one of several neighbourhood drug treatment programmes, as well as to a once-weekly support group offered on-site at the obstetric clinic. The women who attended two or more meetings were classified as attendees. Historically, only 5% of women followed through with formal drug treatment.

The analysis was based on treatment completers. Two groups of pregnant women were compared, support-group attendees and nonattendees. The attendees attended on average 4.9 (mean) and 4 (median) group meetings. The clinical outcomes were:

the estimated gestational age at delivery,

the maternal weight gain during pregnancy,

the infant's birth weight,

the Apgar score,

the incidence of meconium staining,

the incidence of placental abruption,

the rate of self-reported alcohol use at delivery, and

positive urinalysis drug toxicity at delivery.

The attendees and nonattendees had comparable characteristics with respect to demographics, substance use and psychosocial measures. The groups differed in the number of years of education, at the 5% level of significance. The mean number of years of education was 11.2 for the attendees, versus 10.5 for the nonattendees.

Forty-five per cent of the women referred to the support group attended at least two sessions. At delivery, both groups were comparable in the estimated gestational age, and the incidences of placental abruption and meconium staining. The support-group attendees gave birth to infants weighing, on average, 330 g more than infants born to the nonattendee group. One-minute Apgar scores were also higher for the infants of support-group attendees (7.9), compared with those of nonattendees (7.2), p=0.02). No significant differences were observed between the two groups with respect to recent self-reported alcohol use and positive urine drug toxicology.

It was likely that the intervention helped women reduce their overall levels of drug use, although the majority were not successful in achieving total abstinence. On-site support groups may serve as an attractive alternative for pregnant women who are unwilling to enter more formal treatment for their alcohol and drug problems.

The authors did not derive a measure of health benefit. The economic analysis should be considered as a cost-consequences design since the benefits were associated with the effectiveness results.

The direct costs were the medical costs for the mother, from 1 week before delivery to 3 weeks after; and for the infant, from birth to 3 weeks after delivery. These were analysed using the actual charges, as obtained from the Medicaid management information system. The costs were not discounted on account of the short timeframe of the care. The resource quantities and the unit costs were not reported separately. The cost of the support-group programme was not included in the analysis.

The costs were treated in a stochastic way, and the standard deviations (SD) and p-values were reported.

The indirect costs were not evaluated.

US dollars ($).

No sensitivity analysis was performed.

See the 'Effectiveness Results' section.

The mean maternal medical costs were significantly lower for the support-group attendees ($4,095, SD=$1,494) than for the non attendees ($5,016, SD=$3,741), (p<0.01). The average medical costs for the infant, from delivery to 3 weeks postpartum, were significantly lower for the support-group attendees ($1,604, SD=$2,094) than for the non attendees ($3,250, SD=$4,877), (p<0.01).

The costs and benefits were not synthesised.

The economic outcomes associated with the support-group model were encouraging. The improved birth outcomes and the reduced costs (both maternal and infant) were due to the intervention. The support groups were inexpensive to implement and were a way towards the containment of health care resources.

The selection of 'no intervention' as the comparator was justified on the grounds that it represented the usual care.

Evidence was presented to demonstrate to what extent the study sample was representative of the study population. The groups of patients were shown to be relatively comparable at analysis.

The validity of the estimated outcomes may have been limited by the quasi-experimental study design used, and by the presence of selection bias; women who chose to attend the sessions had a better prognosis regardless of the effect of the sessions. Also it was unknown how many sessions the attendees actually attended. It might have been more appropriate to have performed a regression analysis using the number of sessions as a dependant variable.

The authors did not derive a summary measure of health benefit. The analysis was, therefore, of a cost-consequence design.

The costs of establishing the support-group programme were not included in the analysis. The overall costs of maternal and infant care were given, but it was unclear what cost categories were included. This invalidated the costing exercise since the number of sessions each woman attended, and the resource use associated with any improvement in health outcome, were not known. The generalisability of the study's findings, in respect of cost analysis, was also severely limited by the lack of detail about the cost components.

The issue of generalisability to other settings was addressed in terms of the specific components of the programme. In addition, the findings were compared with those from other studies. However, the lack of data on the resource use, including attendance rates, and the unit costs was a major drawback. The authors' conclusions were appropriate for the study population, since there was evidence presented that enabled a comparison of the sample and the population. The authors acknowledged several limitations of their study. IN particular, these included the lack of monitoring drug use, and the possibility of selection bias due to the lack of randomisation.

The authors recommended the use of support groups as a way towards improved health outcomes. They stated that the study supported the association between support groups and clinical outcomes, but did not rule out the influence of the other variables that were not investigated. The findings of the study should be viewed with caution given the lack of cost data, the lack of data on the amount of treatment, and the likelihood of selection bias.

None stated