{
  "id": "2003-february-free-stats-programs-compare-hospitals",
  "slug": "2003-february-free-stats-programs-compare-hospitals",
  "url": "https://dwillis.github.io/uplink/article/2003-february-free-stats-programs-compare-hospitals.html",
  "year": 2003,
  "month": "February",
  "monthNum": 2,
  "headline": "Free stats programs compare hospitals",
  "kicker": "SPOTLIGHT: MEDICINE & HEALTH",
  "author_name": "Andrea Adams",
  "author_title": "computer-assisted reporting specialist",
  "affiliation": "The Fort Worth Star-Telegram",
  "summary": "The Fort Worth Star-Telegram analyzed 5 million hospital discharge records using a federal statistical program to uncover patterns in patient care and hospital quality across North Texas, finding discrepancies in death rates and procedural performance. The newspaper published its findings online so readers could compare hospitals based on specific procedures or diseases.",
  "keywords": [
    "hospitals",
    "analysis",
    "data"
  ],
  "topics": [
    "Data & Databases",
    "Health & Environment",
    "Investigative Methods"
  ],
  "technologies": [
    "SAS",
    "SPSS",
    "FoxPro",
    "SQL Server",
    "Excel",
    "CD-ROM"
  ],
  "full_text": "Analyzing millions of hospital discharge records might seem daunting, but customized statistical programs created for the federal government can churn through huge datasets in just a couple of hours - and help uncover interesting patterns in patient care.\n\nThe Fort Worth Star-Telegram, working with an outside researcher, used one of the programs to help analyze 5 million patient discharge records from the Texas Health Care Information Council for \"Vital Signs,\" a July 2002 special report about hospital quality in Tarrant and Dallas counties.\n\nThe newspaper found discrepancies in death rates for a variety of diseases and in the performance of Caesarian sections. Also, the analysis showed that hospitals using new procedures and medicines were the most effective in helping patients. Many of the dozen stories in the special report focused on specific medical topics, such as cardiac procedures and care for the elderly. The Star-Telegram posted the results of its analysis on the Web so readers could compare hospitals based on specific procedures or diseases.\n\nFirst evaluation\n\nThe Texas Health Care Information Council was created in 1995 to publish report cards about hospitals' performance rates for particular medical services. Nothing had ever been published by the agency. As a result, the Star-Telegram decided to conduct its own analysis to show readers how North Texas hospitals compared.\n\nObtaining the data was easy but costly. The newspaper paid $8,000 for two years of data, which was provided on a CD-ROM for each quarter. Jeff Claassen, the newspaper's computer-assisted reporting specialist, had decided to examine information from 1999 and 2000, the only two full years of data, to decrease the chance that a hospital's ranking might be distorted by an abnormally bad year.\n\nThe Star-Telegram decided to use the same computer program that the Texas Health Care Information Council used to analyze the patient medical procedure data: In-Patient Quality Indicators.\n\nIn-Patient Quality Indicators was created at Stanford University for the Agency for Healthcare Research and Quality, part of the U.S. Department of Health and Human Services. The program comes in two formats: SAS program files and SPSS syntax files. The agency provides the files and documentation for free download on its Web site (www.ahcpr.gov) but users need to have a SAS or SPSS license to run the programs.\n\nRather than obtain a license, the newspaper hired Wayne Lehman, a research scientist at Texas Christian University, to conduct the analysis using SAS.\n\nThe program works by assigning codes to each patient discharge record based on that patient's type and severity of illness. It then compares the patients undergoing a surgery or treatment to a national sample of patients undergoing the same, and adjusts the hospital's mortality rate using the difference between the two. The risk-adjusted rates take into account how old or sick the patients are to make hospital comparisons fairer. The program also calculates the number of times a certain procedure was performed and the rate at which a hospital performed a certain procedure.\n\nBefore analyzing the SAS program results, which were compiled in an ASCII text file, Claassen verified the numbers. He made sure the calculations were done using the entire dataset and checked them against other sources. He focused his analysis on death rates for 10 medical procedures, primarily those involving heart problems and birth. Following advice from a researcher at the agency, Claassen included only those hospitals that performed the procedure at least 60 times during the two years to reduce the risk of unreliable statistics.\n\nClaassen used FoxPro and Microsoft SQL Server to run queries, ranking the hospitals that performed best and worst at each procedure and comparing local hospitals to others in Texas. Charts of these results were compiled in Excel for medical reporter Charlotte Huff, business reporter Trebor Banstetter and Claassen to use. The three reporters called area hospitals with the best and worst rates to find out what kinds of factors can contribute to high or low mortality rates.\n\nClaassen recommends using programs like the In-Patient Quality Indicators because it provides the first widely recognized risk-adjustment methodology, decreasing the chances that a news organization's analysis would be criticized. Claassen said journalists who would like to do a story like this elsewhere should keep in mind that many hospitals are nonprofit organizations and therefore not subject to open-records laws. To get around this limitation, he suggests reading medical journals so you can ask better questions of hospitals with high or low mortality rates.\n\nContact Andrea Adams, Missouri School of Journalism",
  "provenance": {
    "extracted_by": "anthropic/claude-sonnet-4-6",
    "extracted_at": "2026-07-05",
    "verified": true
  },
  "similar": [
    {
      "slug": "1995-april-tracking-hospitals-that-mistreat-babies",
      "score": 0.339
    },
    {
      "slug": "1996-december-rocking-the-rules",
      "score": 0.248
    },
    {
      "slug": "1996-december-diagnosing-hospital-health",
      "score": 0.205
    },
    {
      "slug": "1998-february-health-care-hints",
      "score": 0.204
    },
    {
      "slug": "2000-december-nursing-mistakes",
      "score": 0.198
    }
  ],
  "html_url": "https://dwillis.github.io/uplink/article/2003-february-free-stats-programs-compare-hospitals.html",
  "json_url": "https://dwillis.github.io/uplink/article/2003-february-free-stats-programs-compare-hospitals.json"
}