Online Supplement for: ASSOCIATION OF ASTHMA CONTROL WITH HEALTH CARE UTILIZATION A PROSPECTIVE EVALUATION METHODS Population The Northwest Region of Kaiser Permanente (KPNW) is a large, federally qualified, groupmodel health maintenance organization (HMO) located in Portland, Oregon. KPNW has approximately 450,000 members, whose demographic and socioeconomic characteristics are similar to those of the area population as a whole (E1). The present analysis is based on the results of a survey sent to a subset of KPNW members, aged 18 and older, who had two or more antiasthma medication dispensings in 1996 and/or a hospital or ED visit for asthma in 1994, 1995, or 1996. In addition, all individuals had current KPNW health plan coverage as of June 30, 1997. A total of 13,964 members met these criteria. We surveyed this population between August and September 1997. Everyone received a brief, two-page screening questionnaire (ATAQ, see below). In addition, approximately onequarter also received generic and asthma-specific quality of life questionnaires. Sixty-two percent (8,658) of those we mailed to returned a completed questionnaire, and of these, 60% (5,172) reported that they had a doctor diagnosis of asthma, had taken asthma medications within the past 12 mo, and had answered all the questions relevant for the control index. These constituted the target population for the present analysis. Among these 5,172 individuals, 93% (4,795) had 6 or more mo of KPNW health plan eligibility in the following year, 1998, and thus were included in the analysis. The average duration of health plan eligibility during 1998 for this latter group was 11.8 mo, and 95% had a full 12 mo of eligibility. E1
Survey Instrument The Asthma Therapy Assessment Questionnaire (ATAQ) is a brief, self-administered questionnaire designed to assess level of asthma control and identify possible disease management problems. To assess asthma control, the questionnaire asks about: (1) selfperception of asthma control (did you feel that your asthma was well-controlled?); (2) missed work, school, or normal daily activities due to asthma; (3) nighttime waking due to asthma symptoms; and (4) overuse of quick relief inhaler medication, defined here as more than 12 puffs of a reliever medication on any day in the past 4 wk. With the exception of quick reliever use, which involves two questions, each control dimension is assessed using a single question. Respondents are then graded as either having or not having a control problem in each of these dimensions, and the number of control problems is summed to provide an index ranging from 0 to 4. Although the ATAQ instrument assesses these problems relative to both the last year and the last 4 wk, the analysis reported here is based on 4-wk recall responses. Thus, asthma control as reported in this paper reflects short-term asthma control. Copies of the complete ATAQ instrument and detailed coding instructions are available from Merck & Co. Contact Leona Markson at: leona_markson@merck.com. Health Care Utilization The health care utilization data used for this analysis were derived from a number of large administrative and clinical databases maintained by KPNW and briefly summarized below. Although we define acute care as hospitalizations, ED care, and other acute care provided in the regular outpatient clinics, we recognize that some of this is actually convenience care. Similarly, some of what we call routine care may actually be for an acute exacerbation of symptoms. The distinction, nonetheless, has face validity and should generally be accurate. E2
Inpatient database. Both KPNW and its alliance hospitals use an automated inpatient scheduling system that serves as the basis for the discharge abstract. We counted as inpatient admissions all hospitalizations occurring in 1998 for which the primary discharge diagnosis was asthma (International Classification of Diseases, Ninth Revision [ICD-9] code = 493.x). Emergency department database. Data for ED care are recorded in a separate database from that used for inpatient care. Primary diagnosis is not entered as an ICD-9 code, but instead is captured in an open text field. We modified a previously developed search string that appears to have good sensitivity and specificity for detecting asthma. Only true ED visits were included in the category of ED care. Some visits to the KPNW urgent care clinic are also captured in this database, but were included in the category other acute care. EpicCare database. EpicCare (EPIC) is KPNW s automated clinical information system. This computerized medical record database includes information on all clinic-based outpatient contacts occurring within KPNW. These contacts are all coded using the ICD-9 classification system, although this classification is done by clinic staff using text strings rather than after the fact by trained nosologists. Thus, for example, a physician will select the diagnosis asthma from a drop-down menu, which is then coded internally as 493. EPIC does not formally distinguish between primary and secondary diagnoses. We therefore considered all contacts with an asthma diagnosis in any of the first 10 diagnosis fields to be asthma utilization. Among the visits so identified, asthma was listed as the first diagnosis 63% of the time. Visits conducted by phone were also counted. Outpatient visits for asthma occurring at the urgency care clinic, as well as other outpatient visits at which a patient received E3
nebulizer treatment, were classified as other acute asthma care. All other EPIC contacts were classified as routine care. Claims database. Clinical database (OSCAR) is KPNW s automated claims processing system for recording covered services (generally inpatient and ED care) provided by non-kp facilities. Because any given health care encounter may generate several OSCAR billing records, each with its own primary diagnosis, we classified an encounter as an asthma visit if asthma was listed as the primary diagnosis on any of the 10 largest billings. Outside hospitalizations are also included in the inpatient database; we used OSCAR to capture only outside ED care. Overlapping health care utilization information. We were careful not to overcount encounters. Where such records did exist, we used OSCAR to classify ED care, because OSCAR uses ICD-9 codes versus a text field for the ED database. For other outpatient care, the order of databases from which diagnosis data was taken was EPIC first, inpatient next, and OSCAR last. Outpatient pharmacy database. All medications dispensed at KPNW outpatient pharmacies are recorded in a common database. We used this database to create several summary variables, including (1) total number of antiasthma dispensings (not including prednisone); (2) total number of β-agonist dispensings (inhalers and nebulizers); (3) total number of prednisone dispensings; and (4) the ratio of β -agonist dispensings to inhaled corticosteroid dispensings. Ratios were not calculated for other controller medications, because inhaled corticosteroids were the primary controller therapy used during this time. Although these variables were not used to define any of the health care utilization categories previously described, they were used in some of our prediction equations as noted in the main text. E4
Eligibility Data We used KPNW membership records to compute the number of person-months of eligibility for each individual during 1998. Person-months were then divided by 12 to get person-years, and this information was used to calculate rates of occurrence of various health care utilization outcomes. Only those individuals with at least 6 mo of eligibility in 1998 were included in the analyses. Statistical Methods Rates of health care utilization are expressed per 1,000 person-years of observation and are computed as the total number of events of interest (e.g., hospitalizations), divided by the total number of person-years of observation, X1,000. For these calculations, total events are counted in the numerator, regardless of whether they represent repeat occurrences for given individuals. Proportions of individuals with different types of health care utilization were computed as the number of individuals with such utilization during 1998 divided by the number of individuals. Reference E1. Freeborn DK, Pope CR. Promise and performance in managed care, 1st edition. Johns Hopkins University Press; Baltimore, MD: 1994, p. 1 170. E5