Concept: Families First Screening (FFS) / BabyFirst Screening (BFS) - Method of Creating a FFS / BFS Dataset

Introduction

Families First is a program delivered by Manitoba Child and Youth Programs (formerly Healthy Child Manitoba (HCM)) that offers home visit supports to families from the time they are expecting children to the time their children are about to enter school. Data is collected on the Families First Screening (FFS) form, which includes brief measures of biological, social, and demographic risk factors. Examples of these measures include: low birth weight, alcohol/drug use by the mother during pregnancy, conditions of mental illness, and the mother's level of education completed (less than grade12 or grade 12 and above).

Public Health Nurses in Manitoba attempt to assess all families with newborns within a week of discharge from the hospital. The identification of three or more risk factors indicate that a family may require additional supports such as intensive home visiting, financial support, parenting programs, mental health services, or child care. The FFS, previously known as the BabyFirst Screening (BFS), is the first of two screening stages for Manitoba's Families First home visiting program. The BabyFirst program began on January 1, 2000 and continued until December 31, 2002. On January 1, 2003, the program became known as Families First and continues to provide similar services and programs to families.

This concept provides a brief description of the BFS and FFS datasets and provides links to the HCM forms this data is collected on. The methods section identifies the types of FFS data files relevant to working with this data, identifies some of the variables / indicators available from these data sets, and lists the steps involved in creating the FFS dataset.

NOTE: The method to create a BFS dataset is similar to the FFS dataset, but is not described in detail in this concept. The concept does discuss the BFS data where meaningful, and contains a link to the SAS code for creating a BFS dataset. Step 3 in the Methods section contains a description of the FFS and BFS macro calls, that can be used to create FFS and BFS datasets.

Links throughout the concept provide access to additional information applicable to the BFS / FFS data.

The FFS and BFS Datasets

The MCHP Data Repository contains data from the BFS and FFS programs. As of March 18, 2022 the MCHP Data Repository holds:


• BFS data from January 1, 2000 to December 31, 2002, and
  
• FFS data from January 1, 2003 up to October 12, 2020.

For more information on the BFS and FFS data, please see the Families First Screening (FFS) Data glossary term.

Screening Forms

Data for the BFS and FFS databases are collected on screening forms that are used to assess the risk factors involved. Over time, different versions of these forms have been used. An example of the BabyFirst Screening Form is available on-line in the Next Steps in the Provincial Evaluation of the BabyFirst Program: Measuring Early Impacts on Outcomes Associated with Child Maltreatment (2007) deliverable:


• BabyFirst Screening Form - 2000

Additional examples of the BabyFirst and Families First Screening Forms are available in the Links section below - (internal access only) .

NOTE: The variation in the content/structure of these forms place constraints and restrictions on the use of this data consistently over time. For a description of cautionary notes when working with this data and a list of the differences in the data content/structure between the forms, please read the MCHP internal document titled The BabyFirst Screen (BFS) and Families First Screen (FFS) - Differences between the two screens - (internal access only) - located in the Links section below.

Methods

The following section describes the methods used to create a FFS dataset. This includes a description of the relevant FFS data files and content / indicators and the general steps to follow in order to create a FFS dataset.

NOTE: As mentioned above, the method to create a BFS dataset is similar to the FFS dataset, but is not described in detail in this concept. The focus of this concept is on the FFS data.

MCHP has developed two macros to assist in the creation of a FFS and/or a BFS dataset. For more information on the FFS and BFS macro calls and the required parameters, see step 3 below.

1. The Data Files

There are two types of data files relevant to working with the FFS data:


1. FFS datasets - contain the FFS screening ID (i.e.: bfs_id) and the screening outcomes. There are several FFS data files based on the date of acquisition.
   
   
   • Examples of the outcomes data include indicators for:
     • alcohol and drug use by the mother during her pregnancy;
       
     • maternal smoking during pregnancy;
       
     • parental history regarding anxiety disorders, depression, child abuse, and criminal involvement;
       
     • indication of whether the mother/father completed grade 12; and
       
     • physical and medical characteristics of child.
     
     For a list of specific indicators from MCHP research, please see the section titled Variables / Indicators From the Risk Screening Tool below.
     
     
   • NOTE: Over time, the content/structure of the BFS/FFS datasets has changed. Please read the MCHP internal document titled The BabyFirst Screen (BFS) and Families First Screen (FFS) - Differences between the two screens - (internal access only) - located in the Links section below for a description of cautionary notes when working with this data and a list of the differences in the data content/structure between the forms/datasets.
   
   
2. Linkfile datasets - contain the FFS screening ID (i.e.: bfs_id) associated to a mother and her potential newborn(s).

2. Steps for Creating the FFS Dataset

The steps listed below are a general summary of what needs to be done in order to generate a final FFS dataset. Some details have been omitted. If you would like a more detailed explanation or further your understanding, please refer to the PowerPoint presentation titled Families First Screen (FFS) - SAS Code to Generate the FFS Dataset - (internal access only) - by Wendy Au from the February 27, 2013 MCHP Data Analyst meeting. This concept also contains links to SAS code that have been used for generating the BFS and FFS datasets. Out Of Province (OOP) births are accounted for in the FFS/BFS data/SAS code programs. The FFS SAS code example deals with OOP births for the time period 2008-2010. Prior to this (for 2000-2007), OOP births have already been accounted for. Please see the example FFS and BFS SAS code provided in the SAS code and formats section below - (internal access only) for more detailed technical information.


1. Generate your newborn cohort.
   
   
2. After you have generated your newborn cohort obtain from the REGISTRY file a mom_baby dataset that attempts to link each of the newborns in your newborn cohort to a mom. Previous linkage rates indicate that the mom-baby linkage is close to 99%.
   
   
3. Set all the FFS (social.hcm_FFS_xxxxjan) datasets together. This dataset contains the Families First Screen number (bfs_id) and screening outcomes data.
   
   
4. Set all the Linkfile (social.hcm_FFS_linkfile_xxxxjan) datasets together. This dataset contains the Families First Screen number (bfs_id) and the mom phins and her potential newborns.
   
   
5. Now merge the FFS and Linkfile dataset together by common bfs_id. This dataset is referred to as the FFS_linkfile dataset.
   
   NOTE: In order to maximize the number of newborns screened allow up to a (+/-) 90 day difference in the birthdate value from the two sources.
   
   
6. For each bfs_id record in the FFS_linkfile dataset determine:
   
   1. if the mom phin agrees from the two data sources. If so then mom_match = 1.
      
   2. if the newborn phin agrees from the two data sources. If so then child_match = 1.
   
   
7. Determine the source of the mom phin and baby phin; either FFS, Linkfile or alternative.
   
   
8. Create 4 datasets where each dataset will contain the various combinations of mom phin (if any) and baby phin (if any).
   
   1. FFS_linkfile_1: where mom phin ^= . or baby phin ^= .
      
   2. FFS_linkfile_2: where mom phin from linkfile ^= . or baby phin from linkfile ^= .
      
   3. FFS_linkfile_3: where mom phin from FFS ^= . or baby phin from FFS ^= .
      
   4. FFS_linkfile_4: where an alternative mom phin ^= . or alternative baby phin ^= .
   
   
9. Set all the FFS_linkfile_x (ie. where x = 1, 2, 3 and 4) together. This dataset is referred to as the FFS_linkfile dataset.
   NOTE: There will be multiple repeat bfs_id records, but this is okay as ultimately the dataset is reduced down to one bfs_id per baby and mom pair.
   
   
10. There are some bfs_id(s) where a mom phin exists, but a baby phin is missing. Use the mom_baby dataset obtained from the REGISTRY to potentially identify the baby associated with that mom. This dataset is referred to as the relink_mom_baby dataset.
    
    
11. Set the FFS_linkfile dataset and the relink_mom_baby dataset together.
    
    
12. Create 2 datasets:
    
    1. one_bfs_id dataset (ie. baby/mom record is associated to only 1 bfs_id)
       
    2. one_plus_bfs_id dataset (ie. baby/mom record is associated to possibly 1+ bfs_id and possibly 1+ formdates).
       
       NOTE: in the one_plus_bfs_id dataset remove duplicate records.
    
    
13. Set the one_bfs_id dataset and the one_plus_bfs_id dataset together. This dataset is referred to as the all_FFS dataset and it contains all the possible baby/mom FFS available. Additional cleaning of records is required.
    
    
14. Now merge the newborn dataset to the mom_baby dataset obtained from the REGISTRY by common baby phin. By doing so you have linked the baby to his/her mom. This dataset is referred to as the newb dataset. Add suffix _newb to all the variables in the newborn dataset.
    
    
15. Next merge the newborn dataset to the all_FFS dataset by common baby phin. By doing so you have linked the baby to his/her screen. This dataset is referred to as the newb_FFS dataset. Add the suffix _FFS to all common variables appearing in both the all_FFS and the newb datasets.
    
    
16. Upon merging by the common baby phin check to see if the mom phins agree from the two 2 data sources.
    
    NOTE: There will be some newborns that report 2 different mothers. However, you will only want to associate one mom to a newborn. In the work completed, when such a scenario occurred, choose the mom phin as reported in the REGISTRY file.
    
    
17. Create 2 datasets from the newb_FFS dataset:
    
    1. FFS_exist_0 (i.e.: FFS_exist = 0 -> baby has NO screen available)
       
    2. FFS_exist_1 (i.e.: FFS_exist = 1 -> baby has AT LEAST one screen available)
    
    
18. Where FFS_exist = 1 reduce down to one bfs_id per baby/mom record.
    
    
19. Set the FFS_exist_0 dataset and the FFS_exist_1 dataset together.
    
    NOTE: if a baby is found in both datasets (i.e.: reports no screen and has screen) keep the baby record that reports a screen.
    
    
20. Now you have successfully generated your final FFS dataset.
    
    NOTES:
    • Depending on the set of inclusion/exclusion criteria(s) applied to the newborn cohort, the % of infants screened will vary. However, based on past experience, one should expect around an 80% screening rate.
      
      
    • The method for creating the BFS dataset is similar to the method for creating the FFS dataset. The detailed methods for creating the BFS are not described in this concept. For more detailed information on creating the BFS dataset, please refer to the appropriate SAS code link in the SAS code and formats - (internal access only) section below.

3. FFS and BFS Macro Calls

MCHP has developed SAS macros to assist in the generation of FFS and BFS datasets.

The following macro call from the MCHP Macro Library can be used to create a FFS dataset:

%_FFS (nbbirth_cohort = ,
nbbirthdt_startdt = ,
nbbirthdt_enddt = ,
FFS_output = FFS_dataset);

nbbirth_cohort = The phins of the children in the cohort.
NOTE: the phin of the child must be named scrphin.

nbbirthdt_startdt = The min birthday value of the phins in the cohort.
NOTE: the format of the date must be 01Jan2003.

nbbirthdt_enddt = The max birthday value of the phins in the cohort.
NOTE: the format of the date must be 31Dec2003.

FFS_output = the name of the final FFS dataset - default is FFS_dataset.

The following macro call from the MCHP Macro Library can be used to create a BFS dataset:

%_BFS (nbbirth_cohort = ,
nbbirthdt_startdt =,
nbbirthdt_enddt =,
BFS_output = BFS_dataset);

nbbirth_cohort = The phins of the children in the cohort.
NOTE: the phin of the child must be named scrphin.

nbbirthdt_startdt = The min birthday value of the phins in the cohort.
NOTE: the format of the date must be 01Jan2000.

nbbirthdt_enddt = The max birthday value of the phins in the cohort.
NOTE: the format of the date must be 31Dec2002.

BFS_output = the name of the final BFS dataset.

Data Cautions / Limitations

• The Screening Forms and thus the data collected and the resulting structure of the FFS data change frequently. Prior to use, please check the availability and contents of the data for completeness and consistency.
  
  
• The 2020 screening year is incomplete, hence use with CAUTION or if possible do not include this screening year in your project until the next cut of FFS data comes in to better capture the 2020 FFS year. However, if your project must look at the 2020 screening year then (maybe) only go up to March 31, 2020 and let your PI know that these screening rates will be lower than expected.
  
  
• See the Data Cautions section in the Families First and BabyFirst Screening Data Description for relevant information.
  
  
• See the cautionary notes in the document titled The BabyFirst Screen (BFS) and Families First Screen (FFS) - Differences between the two screens - (internal access only) - located in the Links section below for more information.

Variables / Indicators From the Risk Screening Tool

In the article Unconditional prenatal income supplement and birth outcomes by Brownell et al. (2016), they used some of the variables from the risk screening tool from 2003-2010 as covariates in developing propensity scores and balancing the differences between groups in regression models during the analysis. Most of the variables from the risk screening tool are simple indicators (yes/no). The list of variables from the tool that were used in this research included:


• received screen prenatally (indicative of higher risk families);
  
• alcohol use during pregnancy;
  
• drug use during pregnancy;
  
• family history of disability;
  
• smoked during pregnancy;
  
• mother did not complete high school;
  
• received welfare (income assistance);
  
• single-parent family;
  
• no prenatal care before 6 months;
  
• mother has depression;
  
• mother has anxiety disorder;
  
• mother has schizophrenia;
  
• mother has a mental disability;
  
• antisocial father;
  
• antisocial mother;
  
• current substance abuse by mother;
  
• social isolation;
  
• relationship distress;
  
• violence between parents;
  
• mother abused as a child; and
  
• maternal diabetes.

Two additional covariates used in the study, available from other data sources, included mother's age at first birth and area-level socioeconomic status (SES). For more information, please read the Brownell et al. (2016) abstract available below.