In late 2010, several medical professionals came to The Seattle Times with concerns that a popular painkiller, methadone, was linked to a large number of accidental deaths.

Our initial hurdle was finding a way to track the number and circumstances of these deaths.

We turned first to a database of death certificates maintained by the state Department of Health. The database incorporates a wealth of information about decedents, such as age, occupation, cause of death and related health factors.

Drawing on the methods of medical researchers, we looked for medical code T403, which identifies methadone-related deaths. But that gave us only a partial picture.

We turned to another Health Department dataset, the Literals Database, which captured any notes physicians included on the death certificates. Based on analysis of 376,075 deaths from 2003 through 2010, we identified 2,173 fatalities linked to methadone. We included only deaths designated as accidental. Suicides were excluded.

While death certificates often represent only a best judgment about a cause of death, our determination of methadone-related cases was bolstered in most cases by autopsy findings that were included in the death records.

As a result, many death certificates included notations as to whether other prescription drugs or other substances were involved, or if the person had a history of drug abuse.

Our next hurdle centered on finding a way to map the addresses of the deceased and determine if there was a correlation between deaths and income levels.

We had learned about a state Health Department research report, spanning 2004 through 2007, which concluded that low-income patients were more likely to suffer accidental overdoses.

Using that report as a guide, we analyzed deaths from 2003 through 2010.

Using ArcGIS mapping software, we plotted the addresses for 2,028 of the 2,173 methadone deaths. Some deaths were not included on the map because of inaccurate or missing information.

Next we overlaid census tracts and calculated how many deaths occurred in each. We joined poverty and income data from the U.S. Census Bureau’s American Community Survey and divided the tracts into quintiles, or five equal groups. The top 20 percent of tracts were grouped as “very high” based on median household income; the bottom 20 percent were “very low.” We did a similar analysis substituting poverty for income.

Finally, we calculated a death rate in each quintile and found a disproportionate number of methadone deaths in the poorer tracts. For example, there were 1.4 deaths per 10,000 people in tracts where the median household income was rated “very high.” In the “very low” income tracts the death rate was triple that at 4.9 per 10,000.

In addition to deaths, we tracked methadone-related hospitalizations and associated costs. We sifted about 5 million hospitalizations using a health-department database called the Comprehensive Hospital Abstract Reporting System.

Without naming patients, this database provides a medical profile and breakdown of costs for individuals discharged from Washington hospitals. Like death certificates, diagnosis information is notated by a standardized medical code. We used codes 96502 and E8501 to identify methadone cases.

To track Washington’s rising use of methadone, we relied on a database from the U.S. Drug Enforcement Administration called Automation of Reports and Consolidate Orders System (ARCOS).

Our reporting extended beyond databases. We reviewed dozens of research reports and government studies about methadone and other narcotic prescription drugs.

To research how the state has dealt with complaints about health-care practitioners, we obtained thousands of pages of investigative files. We also interviewed more than 100 people, including state officials, health-care practitioners and patients.