Computerized Reminders to Improve Isolation Rates of Patients with Drug-Resistant Infections: Design and Preliminary Results

Abstract

Infections with multi-drug resistant organisms (MDROs) are a steadily increasing threat to inpatients as they move through the hospital. We constructed a computerized reminder system, integrated with existing data sources maintained by the infection control service. We tested the hypothesis that computerized reminders from a centralized infection control expert system can improve rates of barrier isolation in patients with MDROs. Immediately after institution of this reminder system, isolation of eligible ward patients increased from a baseline rate of 33% to 95% (p<0.0001). The median time to isolation decreased from 16.6 hours to 0 hours, indicating that the majority of patients post-intervention had isolation orders written prior to arrival on the ward. We conclude that a computerized reminder system can dramatically increase the clinical impact of data sources maintained by hospital Infection Control services.

Background

Methicillin-resistant Staphylococcus Aureus (MRSA) has steadily risen in incidence since it was first reported in the 1960s until 2002 when it accounted for 57.1% of all S. Aureus isolates in intensive care units.¹ Patients infected with this and other multi-drug resistant organisms (MDROs) suffer increased morbidity and mortality and have significantly increased lengths of stay and health care costs. Early contact isolation of MDRO infected patients can limit the spread of this pathogen. ²

At most hospitals, an infection control department maintains a list of patients with a history of MDROs, bed control assists in identifying these patients for clinicians, and clinicians place the orders for the patient to be isolated. Once a patient has an infection with an MDRO such as MRSA or Vancomycin-Resistant Enterococcus (VRE), they are considered permanently colonized, pending an often unsuccessful decontamination process.^3,4 Appropriate isolation of patients with MDROs requires the successful execution of a number of steps, and the attention and efforts of several individuals. In any process requiring continued human attention and effective communication of information, there will be errors and oversights⁵ and a high percentage of patients with a history of MDROs may not isolated in a timely manner.

Our institution is similar to most other hospitals in having a human and paper-based system of infection control. The pre-intervention process for appropriate isolation is depicted in Figure 1.

Figure 1.

The isolation process prior to the reminder system.

Obstacles to this highly human and paper-based process include the following:

1. Data availability - Infection control maintains a paper list of MDRO patients requiring isolation. An updated version of this paper list is not always available for Bed Control to appropriately assign patients to rooms.

2. Personnel shortage – Wishard has only one infection control nurse, who must split her time between data collection and “boots on the ground” infection control practice.

3. Communication gap – Bed Control communicates the MDRO status of a patient to the floor nurse when calling report. The physician who actually writes the isolation order is often not informed.

4. Time constraints/Education – Physicians often lack the time to comprehensively search prior microbiology data, which may contain the only reference to MDROs. Even with this knowledge, physician’s may be unaware that a patient with a history of MDROs is likely still infectious (for many MDROs, essentially a permanent state).

Electronic medical record systems can rapidly disseminate critical information to clinicians and improve compliance with guidelines.^5,6 We hypothesized that a clinical reminder system, using centrally updated electronic data, could improve existing isolation practices, increasing the rates of appropriate contact isolation and decreasing the time to isolation for admitted patients.

Methods

Setting

We conducted a prospective “before and after” interventional study. The study took place at Wishard Memorial Hospital, a 264-bed primary care hospital serving an urban population in Indianapolis. We collected baseline data on inpatient isolation orders for approximately four months from November 1, 2004 until the start of the intervention. The intervention started on February 25, 2005. Results reported in this paper include data collected through March 14, 2005 (18 days of intervention). The Institutional Review Board of Indiana University, Purdue University, Indianapolis (IUPUI) approved the study.

The Infection Control service at Wishard Memorial hospital has long maintained their own records of patients with a known history of MDROs. For this study, we focused on the most common MDROs in our hospital, MRSA and VRE. We performed a onetime batch electronic entry of patients requiring contact isolation. As this list requires daily manual upkeep (as new patients with MDROs are discovered), we created an electronic entry form for infection control to maintain their electronic list. Data from this electronic MDRO list was saved into the Regenstrief Medical Record System (RMRS).⁷

Study subjects included physician housestaff and faculty writing orders for hospitalized patients. Physicians routinely enter all hospital orders into the “Gopher” electronic physician order entry system.^6,7,8 During the order entry process, the system provides clinical decision support to physicians using rule-based reminders. We created G-CARE⁹ rules to display suggested orders for barrier isolation for appropriate patients based on the electronic MDRO list (Figure 2). Computer reminders were displayed each time the physician wrote orders for the patient until the patient was either discharged or a barrier isolation order was written. We did not remind physicians if the patient already had an existing orders for barrier isolation. The isolation reminders could be accepted with one keystroke.

Figure 2.

A typical Barrier Isolation suggested order.

Data collection

For this study, patients were eligible for isolation reminders if they were listed as having evidence of MRSA or VRE infection prior to the date of admission as per the electronic MDRO list. Isolation orders with date and time stamps were extracted from the Gopher order entry system and matched to patients by Medical Record Number. We determined the percentage of admitted patients with prior MDROs that physicians ordered into contact isolation, as well as the time to isolation.

As described, the MDRO list is used by Bed Control at the time of admission to identify patients in need of isolation. Consequently, we measured the time to isolation as the difference in time from arrival to a hospital ward bed to the time of isolation order. For analysis, negative times to isolation order (indicating the order was written prior to arrival on the ward) were treated as zeros.

Statistical Analysis

We compared the proportions of patients isolated during the baseline pre-reminder period and the reminder period using the chi-square test. We compared mean times to isolation using the Wilcoxon Rank Sum Test. We used SAS software 8.2 (SAS Institute, Inc., Cary, NC). All tests were one tailed with a p value < 0.05 considered significant.

Results

In the four month baseline period, 109 patients admitted to the hospital were on the infection control MDRO list and were eligible for isolation (Table 1). Of the eligible patients, physicians wrote barrier isolation orders on 36 (33%). The median time to isolation was 16.6 hours. The mean time to isolation was 4.7 days. During the baseline period, only 8% of the eligible patients had barrier isolation orders prior to arrival to the assigned hospital floor bed.

Table 1.

Ordering rates for barrier isolation for eligible patients.

	Pre-Intervention	Intervention	P Value
Eligible Patients Isolated, n (%)	36/109 (33%)	19/20 (95%)	< 0.0001*
Ordered before arrival on ward, n (%)	9/36 (25%)	13/20 (65%)	0.008*
Mean time to isolation, days (SD)	4.68 (9.76)	0.33 (0.84)	<0.0001†
Median Time to Isolation, hours	16.6	0.00	n/a

^*= Chi-Square

^†= Wilcoxon rank sum

During the 18 days post-intervention described in this report, 20 patients were eligible for isolation, which was comparable to the baseline period (approximately one eligible patient per day). After institution of automatic electronic orders, 95% (19/20) of eligible patients admitted to the hospital were ordered into contact isolation. The median time to isolation decreased substantially from 16.6 hours to 0 hours (i.e. the patient had barrier isolation orders by the time of arrival to the assigned hospital floor bed). In 65% of cases (13/20) the clinician wrote isolation orders prior to the patient moving from the Emergency Department to the assigned hospital bed yielding negative or zero times.

In 75% of cases, clinicians accepted the first suggested order. There was one instance where the physicians never accepted the reminder. And, in one other instance, physicians for unclear reasons rejected 14 reminders before finally accepting barrier isolation orders.

Discussion

We found that a computerized clinical reminder system significantly increased the appropriate ordering of contact isolation for inpatients with MDROs, from a baseline rate of 33% to fully 95%. When limited to patients who had isolation orders, we found that the reminder system markedly improved the timeliness of these orders, avoiding a baseline average delay of more than three days, during which the patient might infect other hospitalized patients.

Informatics principles are well suited to streamlining the infection control process. Infection control providers face the daunting task of tracking a patient’s location, infection status, and isolation status, and then delivering appropriate information to the responsible provider at the exactly right point in time – i.e., at time of hospitalization. Through our reminder system, we have freed up the infection control providers to attend to education, active surveillance, and outbreak investigation. Infection Control updates the electronic list of MDRO patients on a daily basis. We have taken advantage of existing infection control workflow, the ubiquity of physician order entry and an EMR to extend the reach of a single infection control provider (Figure 3).

Figure 3.

Extending the reach of a single infection control provider.

By directly alerting the physician to the patient’s MDRO status at the time of initial orders, we were able to greatly increase the likelihood that patients had isolation orders prior to their arrival on the inpatient floor. Early identification and intervention is essential to decreasing the spread of MDROs, as the majority may be acquired in-hospital. In one study, 80% of MRSA episodes were hospital-acquired prior to the institution of infection control measures.¹⁰ In our study, Infection Control entry of a patient’s infection status directly into the EMR ensured that expert opinion was instantly available throughout the inpatient setting. Subsequent simple logical rules guided clinicians to rapid and near-complete patient isolation.

Our study is consistent with others’ findings that reminders can be useful for infection control. In one study, computerized clinical reminders were effective at notifying physicians of MRSA status and reducing the average time to obtaining follow-up cultures from 25 days to 3 days.¹¹ However, in another study, presenting computerized reminders about infections to clinicians did not influence the practice patterns of clinicians.¹² This reminder study offered reminders across a range of possible clinical conditions. By focusing on a single fact, i.e. a patient’s infection status, our intervention markedly improved compliance with infection control policy. This is supported by the infection control literature that well designed medical record systems will help stem the tide of drug resistant infections.¹³

In the past, infection control wrote isolation orders directly for the patients, requesting the treating physician’s approval. This had the opposite of the intended effect, and the number of isolation orders declined, and physicians ignored the orders.

This study has some potential limitations. It is early in the intervention phase, and despite our highly significant findings, it is possible that there may be future decreases in effect. As a “before and after” study, it is subject to variations in practice, although the fact that the baseline period immediately preceded a short intervention period, makes it unlikely that we may be witnessing significant trends. Although we were able to achieve isolation orders for the majority of patients prior to arrival in the hospital wards (negative times), time spent in the emergency room represents an ongoing source of infection transmission and represents an opportunity for further intervention.

Future Directions

We plan to further improve the system by automating the update of MDROs directly from the microbiology system. Infection control does minimal interpretation of MDRO status from the printed microbiology daily reports, and automating this simple information transfer would likely reduce delays and result in significant time savings for the infection control service. With MDRO status now located within the RMRS, we can easily make this information available to clinicians in the outpatient setting. Daily e-mails tracking the isolation status on eligible patients could help focus the educational efforts of the infection control nurse. We have already shown that creation of standing orders can significantly further improve compliance with clinical guidelines.⁸ As a next step we will institute standing orders for barrier isolation for patients with a history of MDROs.