A cohort study for derivation and validation of a clinical prediction scale for hospital-onset Clostridium difficile infection.

Can J Gastroenterol. 2012 Dec;26(12):885-8.

Chandra S, Latt N, Jariwala U, Palabindala V, Thapa R, Alamelumangapuram CB, Noel M, Marur S, Jani N.

Department of Internal Medicine, Greater Baltimore Medical Center, Baltimore, Maryland 21204, USA. subhash.budania@gmail.com

Abstract

OBJECTIVE: To develop and validate a clinical prediction scale for hospital-onset Clostridium difficile infection (CDI).

METHODS: A community-based, 360-bed hospital located in the suburbs of a metropolitan area in the United States served as the setting for the present retrospective cohort study. The cohort consisted of patients admitted to the adult medical service over a six-year period from October 2005 to September 2011. The cohort was divided into derivation (October 2005 to September 2009) and validation (October 2009 to September 2011) groups. The primary outcome measure was hospital-onset CDIs identified as stool positive for C difficile after 48 h of hospital admission ordered for new-onset unformed stool by the treating physician.

RESULTS: In the derivation phase, 35,588 patients were admitted to the medical service and 21,541 stayed in hospital beyond 48 h. A total of 266 cases of CDI were identified, 121 of which were hospital onset. The developed clinical prediction scale included the onset of unformed stool (5 points), length of hospital stay beyond seven days (4 points), age >65 years (3 points), long-term care facility residence (2 points), high-risk antibiotic use (1 point) and hypoalbuminemia (1 point). The scale had an area under the receiver operating curve (AUC) of 0.93 (95% CI 0.82 to 0.94) in predicting hospital-onset CDI, with a sensitivity of 0.94 (95% CI 0.88 to 0.97) and a specificity of 0.80 (95% CI 0.79 to 0.80) at a cut-off score of 9 on the scale. During the validation phase, 16,477 patients were admitted, of whom 10,793 stayed beyond 48 h and 58 acquired CDI during hospitalization. The predictive performance of the score was maintained in the validation cohort (AUC 0.95 [95% CI 0.93 to 0.96]) and the goodness-to-fit model demonstrated good calibration.

CONCLUSION: The authors developed and validated a simple clinical prediction scale for hospital-onset CDI. This score can be used for periodical evaluation of hospitalized patients for early initiation of contact precautions and empirical treatment once it is validated externally in a prospective manner.

PMID: 23248788

 

Supplement

Management of CDI in hospital needs a multidisciplinary approach which begins with infection prevention. Routine screening for CDI is no recommended.4 Henceforth, identifying at-risk inpatient population is important. The risk stratification also enables the appropriate use of prophylactic and empirical therapeutic interventions, thus potentially reducing mortality, morbidity and healthcare costs.

A clinical prediction rules (CPR) with good discrimination can identify these at risk patients for hospital-onset CDI. A focused intervention in this high risk population is cost-effective. The CPR described above has following key strengths. First, it has a better discrimination than previously reported CPR for hospital-onset CDI. 10 Second, the components of the scale are readily available from the clinical history except for hypoalbuminemia. Serum albumin is part of the comprehensive metabolic profile which is obtained in most of the patients with longer than 48-hour hospitalization. Third, all the components of the scale are objective and easily reproducible between care providers ensuring consistency in the scale calculation.

In clinical decision rules or prediction scales, the accuracy lies in the details. The scale with multiple variables tends to have better prognostic accuracy. Increase in the number of variables adds complexity in calculation. Implementation of a complex scale in clinical practice is challenging. One way to get around this problem is to have a calculated score auto-generated via the electronic medical record. Development of automatic tools has been described before. These tools retrieve data from the electronic medical records in real-time and provide computed scores for care providers.11 These efforts can facilitate the use of complex scoring systems by making them readily available but are again limited by the fact that comprehensive electronic health records are not available across the country. Also, if a scoring system has subjective components, development of a completely automatic tool is not possible. Henceforth, simplicity of the scale and ease in calculation are essential for acceptance of a clinical prediction scale in clinical practice.

Objectivity in the CPR components mentioned here enables development of an automatic tool which can calculate the scale for the providers. Figure 1 demonstrates schematic work flow of incorporating the CPR in to clinical practice.

 

Subhash Chandra-1

 

Multiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier SchönmannMultiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier Schönmann