A prospective study was performed from July to August 2010 at Rush University Medical Center, a 676-bed urban academic medical center, as a part of a hospital quality improvement project to accurately identify patients admitted from other health care facilities for multidrug-resistant organism active surveillance. All adult (≥18 years old) inpatients, excluding pregnant and postpartum women, were eligible for inclusion in the study. Three to 4 wards were selected randomly each day, and all patients on the wards who were present in their room at the time of interview were invited to participate in the study. Each patient was included only once. Patients were excluded if they declined to participate or if they were incapable of participating in the interview and a family member was unable to be reached. This project was determined by the Institutional Review Board of Rush University Medical Center to be exempt from review.

Of 599 eligible patients, 523 (87.3%) were included in the study. Of the 76 excluded patients, 61 were away from their room at the time of interview; the remainder was excluded because of patient declination to participate (n = 6) inability to reach a family member (n = 8) or family member unawareness regarding patient’s residence (n = 1). Overall, 85 (16.3%) of 523 patients were admitted from another health care facility, with 13 patients admitted from a long-term care facility and 72 patients admitted from other acute care facilities. In detecting patients admitted from another health care facility, the point of origin code had a sensitivity of 86% (95% confidence interval [CI]: 77-92) and a specificity of 98% (95% CI: 97-99); the κ value was 0.86 (95% CI: 0.80-0.92). The physician-documented history had a sensitivity of 75% (95% CI: 65-84), a specificity of 98% (95% CI: 96-99), and a κ value of 0.77 (95% CI: 0.70-0.85). Positive predictive value and negative predictive value were similar between the point of origin code and physician-documented history (Table 2).

In detecting patients admitted from long-term care facilities, the point of origin code had a sensitivity of 50% (95% CI: 23-77), compared with a sensitivity of 71% (95% CI: 42-92) for the physician-documented history. Specificity was similar between the 2 groups. In detecting patients admitted from acute care facilities, the point of origin code had a sensitivity of 93% (95% CI: 84-98), and the physician-documented history had a sensitivity of 76% (95% CI: 64-85), with similar specificity between the 2 groups.

Of the 14 patients admitted from long-term care facilities, 7 were not detected by the point of origin code in our study. Further investigation of their medical records revealed the reasons for the 7 failed identifications: 3 patients were scheduled admissions for surgery and were transported to the hospital by family members. Three additional patients were directly admitted from physicians’ offices and also transported to the hospital by family members. None of these 6 patients had a face sheet from their long-term care facility in their charts. One remaining patient was misclassified because of error by admitting personnel.

The age, gender, and proportion with Medicare or Medicaid insurance were similar for health care facility patients and non-health care facility patients (all P > .05, data not shown). Over twice as many health care facility patients were in contact isolation for any reason compared with non-health care facility patients (15.3% vs 7.1%, respectively, P = .01). A 3-fold greater proportion of health care facility patients were nonverbal or had an altered mental status compared with non-health care facility patients (32.9% vs 9.1%, respectively, P < .001). Health care facility patients were also twice as likely to be admitted to an intensive care unit compared with non-health care facility patients (31.8% vs 16.4%, respectively, P = .001).

We developed a strategy of using a point of origin code from a widely utilized billing form (UB-04) to efficiently identify patients admitted to our hospital with an immediate prior stay in another health care facility. Using a gold standard of patient self-report, the point of origin code had excellent sensitivity and specificity in identifying patients admitted from other health care facilities and performed comparably with manual review of physician-documented history. The point of origin code performed well for identifying patients admitted from acute care facilities but had poorer sensitivity in identifying patients admitted from long-term care facilities.

Our findings suggest that determination of prior health care facility exposure can be potentially automated using existing administrative data. Automated tools for infection control activities have become increasingly widespread in recent years. For example, they have been used to alert physicians when a patient with a multidrug-resistant organism is readmitted to the hospital, calculate device-days for reporting of health care-associated infections, and detect potential outbreaks.^7-10 Automated methods have been shown to be as accurate as manual methods, with increased efficiency and timeliness.⁹ Staff time saved by using automation may be used by infection preventionists to focus on prevention rather than perform surveillance.¹¹

The point of origin code is typically obtained within 24 hours of a patient’s admission and can be used to generate an electronic flag available to hospital personnel. The flag would be useful to infection preventionists interested in rapid identification of patients admitted from other health care facilities, especially in regions where long-term care facilities and long-term acute care hospitals are significant reservoirs for highly resistant organisms such as carbapenem-resistant Enterobacteriaceae.^12,13 Such a flag could also trigger automated orders such as targeted active surveillance cultures or isolation precautions. Because all US hospitals that receive reimbursement from the Centers for Medicare and Medicaid Services are already required to collect point of origin code information at the time of admission for billing purposes, the point of origin code can be readily utilized, with or without an existing electronic medical record. The point of origin code has the distinct advantage of being standardized; it does not require manual chart review of physician documentation, which can be labor intensive for infection preventionists. Furthermore, health care facilities have a strong incentive to correctly document administrative data because it has a direct impact on their financial reimbursement. Nurses and physicians, on the other hand, may not complete these sections of the note or include updated information because they may not be required to do so.

Both the point of origin code and the physician-documented history were not sensitive in identifying long-term care facility patients compared with patients from other health care facilities. This likely reflects the more diverse points of entry of long-term care facility patients who present to our institution compared with patients admitted from acute care facilities. Long-term care facility patients who are brought to the hospital by family members directly from outpatient clinics or for a scheduled procedure, for example, receive the default designation of a non-health care facility source because they sometimes arrive without identifying documentation (such as ambulance paperwork) from their institution. This could be improved by having registration personnel ask each patient or family member a similar question to what we asked in our study: “Where were you staying immediately before being admitted to Rush University Medical Center?” A significant proportion of patients from long-term care facilities in our study (32.9%) were nonverbal and unable to provide this information, but we suspect that all of these patients were brought to the hospital by ambulance personnel (and therefore accompanied by an ambulance report form and/or face sheet) or by a family member. Of the 7 long-term care facility patients who were incorrectly classified as “non-health care facility” by the point of origin code, 6 were brought to the hospital by a family member and could have been correctly identified as facility patients if registration personnel interviewed the patient’s family.

A limitation of our study was the small sample size of long-term care facility patients (n = 14), which contributed to the poor sensitivity and wide confidence interval of the point of origin code in detecting patients from long-term care facilities. Further studies should be performed to assess the sensitivity of the point of origin code in this specific population. Another limitation of our study is that it was performed at a single site; the performance of the point of origin code may vary from one hospital to another depending on local strategies in collecting the point of origin code information. Theoretically, point of origin information may be collected > 24 hours following admission because the UB-04 form is not submitted until the patient is discharged. However, because accurate documentation of the point of origin code affects hospital reimbursement, we believe that hospitals have similar incentives to optimize collection of point of origin code information in a timely manner, and the most opportune time to interview the patient is at the time of admission. A third limitation is that our tertiary academic institution may admit a patient population that differs from others in the prevalence of admissions from other health care facilities. Although variation in the prevalence of prior health care facility types would alter the positive and negative predictive values of the point of origin code from one institution to another, sensitivity and specificity are intrinsic to the point of origin code itself and are not affected by prevalence.

The UB-04-based point of origin code is an accurate and efficient tool to identify patients admitted from other health care facilities. Such information can increase infection control situational awareness and allow identification of patients at risk for carriage of multidrug-resistant organisms.