Data from the ED component of the National Hospital Ambulatory Medical Care Survey (NHAMCS) from 1993–2003 were combined for analysis [5-15]. The NHAMCS is a 4-stage probability sample of visits to non-institutional general and short stay hospitals, excluding Federal, military, and Veterans Administration hospitals, located in the 50 States and the District of Columbia [16,17]. The NHAMCS is conducted annually and covers geographic primary sampling units, hospitals within primary sampling units, EDs within hospitals, and patients within EDs. Trained hospital staff collect data during a randomly assigned 4-week data period for each sampled hospital, approximately once every 15 months [18]. Review of data collection is performed by a U.S. Bureau of Census field supervisor. Quality control includes computer checks to assess inconsistencies with value ranges, a two-way 10-percent independent procedure for medical and drug coding, and adjudication by the National Center for Health Statistics (NCHS) for ambiguous or illegible responses for fields including reasons for visit and diagnosis. The non-response rate for most items is less than 5%, and error rates are less than 2% for items that require medical coding [17]. When the data collection forms are completed, they are sent to Constella Group Inc., Durham, North Carolina where they are coded by experienced personnel using the International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9-CM). National estimates are obtained through use of assigned patient visit weights and are rounded to the nearest thousand. A multistage estimation procedure consists of inflation by reciprocals of the sampling selection probabilities, adjustment for non-response, and a population weighting ratio adjustment. The NHAMCS data form is devoid of patient identifying characteristics. A more detailed description of the NHAMCS data collection and estimation procedures is available for review at the National Hospital Ambulatory Medical Care Survey (NHAMCS) Public-use Data Files web page and in the technical notes section of each year's NHAMCS Emergency Department Summary [19,20].
The NHAMCS allows for specification of up to three physician diagnoses. In this study, we first identified all visit records with the physician diagnosis ICD-9-CM 577.0 (acute pancreatitis) in any position and determined estimated ED visits and visit rates. For all subsequent analysis we then excluded visit records if ICD-9-CM 577.0 appeared only as the third-listed diagnosis or if ICD-9-CM 577.1 (chronic pancreatitis) appeared as a physician diagnosis in any diagnostic position. This was done to minimize possible problems from miscoding of chronic pancreatitis exacerbations as episodes of acute pancreatitis, and to eliminate equivocal diagnoses of acute pancreatitis.
We examined ED visits by patient age, sex, race, Hispanic ethnicity, and insurance status and by hospital metropolitan statistical area (MSA) status and region (Northeast, Midwest, South and West). MSA and U.S. region categories represent standardized geographical divisions defined by the U.S. Bureau of the Census; essentially, a hospital in a MSA is urban [21,22]. Hispanic ethnicity was not imputed by the NHAMCS from 1997–2002, and thus data on Hispanic ethnicity are not available for those years. U.S. visit rates were computed using mid-year age, sex, race, ethnicity, and metropolitan status specific population estimates from the U.S. Census Bureau; all rates were reported per 10,000 individuals per year for the U.S. resident population. Investigators also re-examined primary results for visit rates using the civilian population, as recommended by NCHS, and results were similar (data not shown). Overall average annual rates for the entire study period, where reported (such as in Table 1), were calculated by dividing the total number of estimated ED visits by the sum of the midyear estimates for each of the eleven years of the study. ED management focused on medications, diagnostic imaging, and ED disposition (e.g., hospital admission). Disease severity data are limited but include urgency at triage. To keep analyses between earlier and later years consistent, we coded visits that occurred after a change in coding in 1997 (1997–2003), as "urgent/emergent" if immediacy to be seen was recorded as "less than 15 minutes" or "15–60 minutes," and as "non-urgent" if recorded as ">1–2 hours" or longer. From 1993–94 up to five medications were recorded per visit, from 1995–2002 up to six medications were recorded per encounter, and in 2003 up to eight medications were recorded per encounter, with medications coded as per published NCHS definitions [23-25]. All recorded medications were considered for analysis. Therapeutic class of medication (eg. "antibiotic" or "analgesic") was based on the National Drug Code Directory.
We performed data management and analysis using STATA 9.0 (StataCorp, College Station, TX). A masked ultimate cluster sample design was used to estimate variance. NCHS considers an estimate to be unreliable if it has a relative standard error (SE) of more than 30%. In addition, estimates based on fewer than 30 records are considered inherently unreliable, regardless of their SE. For the current analysis, we determined point estimates and 95% confidence intervals (CIs) for ED visits – both absolute numbers and population rates – as well as for visits by patient, age, sex, race, ethnicity, geographic region, and MSA status. Visits also were analyzed for frequency of hospital admission, analgesic administration (including narcotics), computed tomography (CT) or magnetic resonance imaging (MRI) use, and ultrasound use. CT, MRI, and ultrasound use were not recorded in 1993–94, so these results describe practice patterns from 1995–2003. Pearson's chi-square was used to assess differences between groups. Chi-square for trend was used to evaluate trends over the 11-year period. Visits were combined into 2-year groups (with the exception of 2003) to determine whether trends existed for total visits, or by sex or race. Multivariate logistic regression was used to evaluate independent predictors of hospital admission. Two-sided p-values of less than 0.05 were considered statistically significant. Our study was conducted with the approval of the Massachusetts General Hospital Institutional Review Board.
Answers to all your Biology Questions
