An Electronic Medical Record Alert of Progesterone

Preterm birth affects a significant number of babies born every year, increasing risks of death and life-long disability. Progesterone has been found to halve the odds of preterm birth <34 weeks and neonatal death. Although progesterone is a proven standard medication for pregnant women at risk of preterm birth based on some guidelines (e.g. recommendations by the Current Society for Maternal-Fetal Health Medicine, and National Institute for Health and Care Excellence), few women are receiving this effective prevention. There is an urgent need to improve prevention of preterm birth with progesterone, which the investigators propose to do with an innovative and cutting-edge best-practice alert in electronic medical records (EMRs). Improving care may decrease the risks related to preterm birth, and improve patient experiences and outcomes for women and their babies.

This study will test the feasibility of an EMR alert for progesterone, which provides the opportunity for researchers and care providers to implement and advance the use of proven medical practices in preventing preterm birth and decreasing associated risks of death, life-long disability, family stress, health care use, and lost economic potential. The EMR alert would be transferable to other institutions or regions of the province, and is low-cost, sustainable, embeddable and scalable in the current system health system.

STUDY DESIGN:

The investigators propose a pilot cluster randomized control trial (RCT) with a 1:1 allocation of 8 clinics randomized to usage of an EMR alert (intervention) versus usual prenatal care without the usage of the EMR alert (control). Individual women will not be approached since this is a pilot cluster RCT. Rather, clinics (and care providers therein) have been approached and have agreed to participate in this study.

Investigators have followed the SPIRIT statement (Standard Protocol Items: Recommendations for Interventional Trials).

RANDOMIZATION:

The unit of randomization will be the clinic providing pregnancy care. All pregnant women receiving care at a clinic randomized to the intervention will receive the intervention, while all pregnant women receiving care at a clinic randomized to usual care will receive usual prenatal care. Investigators used a 1:1 allocation between the intervention and the control groups.

STUDY PROCESS:

This study involves a minimal risk intervention, as it is a pilot cluster RCT of the implementation of an EMR alert (regarding progesterone prescription for care of pregnant women at risk of preterm birth, a practice considered standard of care in many guidelines). The Hamilton Integrated Research Ethics Board (HiREB) has approved for clinics to be randomized and all eligible women at intervention clinics to receive the intervention (and all eligible women at control clinics to receive routine prenatal care). As this is a minimal risk intervention and a pilot cluster RCT, individual participant consent is not required. Rather, informed, written consent of the clinics and care providers at the clinics will be obtained for their participation in this study and the use of their questionnaire and structured interview responses. There will be no interim analysis given the minimal-risk nature of the intervention, and the pilot nature of the study, and hence investigators will not have a Data Safety Monitoring Board (DSMB), but will have a Steering Committee.

DATA COLLECTION:

Data will be collected at the end of pregnancy from the:

1. 3-page Perinatal Records mandated by the Ministry of Health: baseline characteristics*, process outcomes;
2. Care provider surveys and structured interviews (in intervention group): feasibility, provider outcomes.

• Baseline characteristics will include: maternal age, education level, ethnic/racial background, pregnancy history (gravidity, parity), gestational age upon first visit to randomized clinic, pre-pregnancy body mass index, chronic health conditions, smoking/alcohol/street drug use during pregnancy, and rate of short cervix.

SAMPLE SIZE:

Investigators have provided a sample size justification, rather than a calculation, for the following reasons:

1. It is recognized that 'in general, sample size calculations may not be required for some pilot studies'
2. Given that this is a feasibility study, it was not designed to have statistical power to detect a difference between the 2 treatment groups.
3. The size of the intraclass correlation coefficient (ICC) required for a sample size calculation is currently unknown.

The sample size was based on feasibility considerations as follows: In the 8 clinics, there would be approximately 2400 women over the year. Based on a chart audit, the investigators estimate that 8% of these women (approximately 192 women) would be available for exposure to the intervention.

STATISTICAL ANALYSES:

Analysis will be done at the patient level with the exception of care provider outcomes. Baseline characteristics will be compared between women in the intervention clinics versus those in the control clinics. Continuous data will be compared using t tests for means (standard deviations) or Mann-Whitney for medians (interquartile range), as appropriate. Proportions will be compared using Chi-squared tests.

The analysis of the primary outcome, feasibility, will be based on descriptive statistics of the proportions (%) of clinics that successfully apply the alert (i.e. get it set up in their EMR), and of care providers would recommend the alert to colleagues; as well as the proportions (%) of approached clinics that agree to randomization, and have completeness of outcome data.

Secondary outcomes (process and care provider outcomes) will be analyzed using t tests or Chi-squared tests comparing the intervention group versus the control group. The investigators will use intention to treat analysis: i.e. outcomes of all eligible women in the intervention group, whether they received the intervention or not, will be evaluated within the intervention group.

Two sensitivity analyses will be done: 1) a "per-protocol analysis", comparing women who were prescribed progesterone (either in the intervention or in the control group) to those who were not prescribed progesterone (either in the intervention or in the control group); 2) comparing results in women with and without complete data.

The investigators will control for potential covariates which may not be evenly distributed between the intervention and control groups (e.g. age, socioeconomic status, etc.). Since observations within each participating clinic will be assumed more likely to be similar than observations between clinics, a logistic model using a conditional (for paired data) generalized estimating equation (GEE) method will be performed to account for this clustering effect within clinics, incorporating both within-clinic and between-clinic variations. An intracluster correlation coefficient (ICC) and variance inflation factor (VIF) will also be calculated to assess the impact of the clustering effect.

Results will be considered statistically significant at two-sided alpha of 0.05. A modified Bonferroni correction will be used given the multiple secondary outcomes. Analyses will be performed using SAS-PC statistical software (version 9.2; SAS institute Inc., Cary, NC).

TEAM:

Principle Investigator: Sarah McDonald, MD, MSc (Clinical Epidemiology), FRCSC, is an Obstetrician, Professor in the Department of Obstetrics and Gynecology at McMaster, and a Tier II Canada Research Chair.

Co-Investigators:

Lehana Thabane, PhD, is a statistician/RCT expert and is the Director of the Biostatistics Unit at the Centre for Evaluation Medicine at McMaster and the Associate Chair of the Department of Clinical Epidemiology and Biostatistics.

Prakesh Shah, MD, MSc, is a neonatologist and Professor in Paediatrics at the University of Toronto.

Karim Keshavjee, CCFP, MBA, MD, MSc, is a practicing clinical information technology architect, and an Adjunct Professor with the Institute of Health Policy, Management and Evaluation at the University of Toronto.

Kathryn May, JD, is an Ministry of Health and Long-Term Care Program Analyst responsible for EMR-related files, eHealth Strategy and Investment Branch, and is responsible for the strategy, funding and oversight of clinician eHealth.

Collaborators:

Kate Robson is our patient representative.

Care providers whose clinics would be involved have provided letters of support. Those whose clinics are randomized to the intervention will give input into the 'alert', after which it will be further revised. All clinicians will be involved in interpretation of the results.