The goal of the proposed project is to develop computerized algorithms to screen hospitals for potentially substandard care using readily available administrative data pertaining to the inpatient stay and outpatient and other nonacute care in the period immediately following discharge. This algorithm will build from the Complications Screening Program (CSP), a computerized method for identifying possible in-hospital complications using discharge abstract data and developed with federal funding (AHCPR HS 06512, 9/30/90-9/29/92). The proposed research recognizes that many complications of hospital Care, such as nosocomial infections, are not detected until the patient leaves the hospital. This is especially likely as lengths of stay continue to decline and services are increasingly pushed into outpatient settings. Therefore, this outpatient (O) version of the CSP (or CSP-O) considers information on outpatient diagnoses and events immediately after discharge (e.g., procedures, nursing home admissions). Specific project objectives are twofold. The first is to develop the CSP- O, as follows: (a) create the clinical logic by designating diagnoses or patterns of service use in the immediate postdischarge period that may reflect complications relating to prior inpatient care; (b) devise an approach to adjust for patients' a priori clinical risks of poor outcomes and complications by focusing on coexisting chronic conditions; and (c) create a computerized algorithm to represent the clinical logic as closely as possible using readily available administrative data. The second is to examine the results of applying the CSP-O to a Medicare Part A and B administrative data file including all elderly Medicare admissions to a stratified random sample of 500 acute care hospitals. The data file will contain discharge abstract information linked to postdischarge ambulatory and other data (e.g., nursing home, home health agency). Analyses will focus on the following: (a) documenting the yield of the various screens (i.e., rates at which cases are flagged as potential complications) across hospitals and at the individual hospital level; (b) exploring which patient-level factors are related to particular complications (e.g., chronic conditions, admission source); (c) examining which hospital-level factors (e.g., bed size, teaching status, location, ownership, surgical volume) are related to higher-than-expected complications rates; and (d) Comparing resource consumption (e.g., ambulatory services, readmissions) of cases with complications to that of other patients. The CSP-O could be a tool for research studies examining hospital outcomes using large administrative databases a way to assist in targeting more efficiently the expensive chart-based reviews of quality and patient outcomes. We recognize, however, the serious limitations of administrative data, especially that pertaining to outpatient care. The CSP-O will require validation as an indicator of in-hospital quality of care.
