About the Data
PublicSource made its first request for information regarding medications prescribed to juvenile offenders in the six state youth correctional facilities in October 2013.
The Pennsylvania Department of Human Services denied the requests. The department said the medication information was kept in individual juvenile files and invoices that contained private health information. That meant the files were not public records, they said, and there was no requirement to redact the identifying information.
The department also refused to share the names of the doctors, citing several exemptions in the state's open-records law but stressing that the release of the doctors' names would put them in danger.
PublicSource appealed to the state Office of Open Records, which agreed with the denial of medications information. The office did order the release of the names of doctors who were under contract to provide services in the state-operated youth development centers and youth forestry camps.
The department appealed to the Commonwealth Court of Pennsylvania, which selected the case for mediation. Before mediation, the department called PublicSource to negotiate.
An agreement was struck for PublicSource to withdraw its request for the doctors' names in exchange for seven years of invoices that show the medications ordered by the facilities.
In July 2014, PublicSource received a bankers box of about 4,000 pages of monthly invoices for each of the six facilities for the years 2007 through 2013.
The information on the scanned paper records was entered by freelancers into a database structure created by PublicSource. The digitized records were then sent to a data entry firm to audit the database for accuracy. PublicSource also conducted integrity checks on the data throughout the process.
The database includes drug name, drug dosage, the unit (such as mg, ml), medicine form of the drug (such as tablet, injection), which facility ordered it, how many units were ordered and the cost of the drug purchase.
We used a tool developed by the National Library of Medicine, called RxMix, to standardize drug names and classify into type of medication (antipsychotic, antacid, anti-inflammatory, etc.).
Focus and methodology
The five classes we focused on were: antipsychotics, antidepressants, mood stabilizers, anti-ADHD and anti-anxiety medications.
We performed additional data quality checks after running the data through the RxMix tool.
We removed certain non-drug entries that produced errors. For instance, purchases like an "Ace bandage," a simple elastic bandage used to restrict the movement of a sprained ankle, was misidentified as a "transdermal patch," an adhesive patch that, when applied to the skin, delivers drugs into the system.
We also removed any records that did not have enough information to be processed in RxMix.
With the help of the Institute for Nonprofit News, we used a program called Django as an interface for analyzing and eventually displaying our data.
PublicSource consulted with two leaders in the field of psychotropic medication use in children and adolescents to develop our methodology. They were:
- Dr. Mark Olfson, professor of psychiatry at Columbia University Medical Center and a research psychiatrist for the New York State Psychiatric Institute, and
- Dr. Christoph Correll, a professor of psychiatry at Hofstra University's School of Medicine and medical director of the recognition and prevention program at the Zucker Hillside Hospital.
At their advice, we treated purchases as monthly prescriptions. We compared this to the average occupancy over time, provided by the Department of Human Services, for each facility.
We used polypharmacy rates found among the state's foster children by PolicyLab of the Children's Hospital of Philadelphia - and supported by journal studies and medical experts - to account for when a juvenile might be on two antipsychotics or two antidepressants.
However, the analysis could not estimate polypharmacy across drug types. For instance, it is common for an antipsychotic to be added to an antidepressant to improve its efficacy.
Applying the polypharmacy rates reduced the percentage of youth offenders who could be treated at any given time. For instance, the antipsychotic prescriptions would have been sufficient to treat an average of 38 percent of the juveniles before the adjustment for polypharmacy. When we factored in juveniles who could be prescribed more than one antipsychotic, the average dropped to 33 percent.
To better understand our findings and to continue to ask important questions of the data, PublicSource also conducted about 40 interviews with judges and attorneys, medical experts and researchers, advocates, and state officials.
While PublicSource attempted to obtain the most accurate data it could on the medications prescribed to juveniles in the youth development centers and forestry camps, there are some important caveats about the findings.
Misclassifications: The RxMix API was developed by the National Library of Medicine to be used as a tool to normalize and classify drugs. The system takes in dirty information - misspellings, abbreviations, improper dosages - and attempts to output cleaned, normalized, accurate drug information. Every attempt was made to ensure proper RxMix processing, but there may be certain records that were misclassified.
Assumptions about prescribing: Due to health privacy laws, the medications were not linked to any particular case or individual. PublicSource had to make certain educated assumptions about the prescribing habits at the facilities, which we detail in the methodology above. We made every effort to report a conservative estimate of psychotropic medication use.
PublicSource would like to thank:
- Drs. Olfson and Correll for sharing their time and medical expertise;
- Olivier Bodenreider at the National Library of Medicine, for his assistance with the tool that allowed us to actually use the data, and;
- Ryan Nagel, Adam Schweigert, Ben Keith, Kaeti Hinck and Dani Litovsky at the Institute for Nonprofit News for their help in designing and developing the app.
Contact Halle Stockton at firstname.lastname@example.org or 412-315-0263.
Stories by Halle Stockton. Data analysis by Alexandra Kanik, Eric Holmberg and Stockton. App development by Kanik with help from the Institute for Nonprofit News. Graphics and charts by Kanik. Videos by Ryan Loew. Illustrations by Anita DuFalla.