Journal watch: Recent studies of note

Bedside tool to predict in-hospital death after stroke, and more.

Bedside tool predicts in-hospital mortality after ischemic stroke

Researchers have developed a new bedside tool to determine death risk for acute ischemic stroke inpatients that aims to be more practical and user-friendly than existing prediction models.

Other models haven't been well incorporated into clinical practice, in part because they involve data that are not routinely collected or because they are time-consuming, the researchers noted. For their model, the researchers used the Get With the Guidelines-Stroke (GWTG-Stroke) data registry to analyze 274,988 ischemic stroke admissions from 1,036 hospitals that occurred between October 2001 and December 2007. They randomly divided the sample into derivation (60%) and validation (40%) samples, and used logistic regression to determine independent mortality predictors and assign point scores for a prediction model. The model uses patient demographics, medical history, mode of presentation to the hospital (i.e., by ambulance) and timing of admission to predict a patient's risk of dying in the hospital after admission for a stroke. For the 40% of patients who had a National Institutes of Health Stroke Scale (NIHSS) recorded, they separately derived and validated a risk score.

Of patients admitted for ischemic stroke, 5.5% died in the hospital. Characteristics associated with in-hospital mortality included older age, mode of arrival history of atrial fibrillation, dyslipidemia, previous stroke and/or myocardial infarction, carotid stenosis, diabetes mellitus, peripheral vascular disease, current smoking and weekend or night admission. There was very high correlation in the validation sample for observed versus predicted mortality (r2=0.995), indicating excellent calibration. Model discrimination was quantified by calculating the C statistic from the validation sample, which was 0.72 in the overall validation sample (i.e., the GWTG-Stroke tool), 0.85 in the model that included the NIHSS score, and 0.83 in the model that used only the NIHSS score. Thus, the NIHSS score alone was a very strong predictor of mortality risk, showing the relationship between stroke severity and short-term mortality. Results were published Oct. 12, 2010 in Circulation.

The GWTG-Stroke model doesn't include information related to in-hospital care or medical complications, the authors noted. “This model therefore predicts the expected in-hospital mortality at the time of presentation, and as such, the predictions could be used to guide subsequent in-hospital medical care,” they wrote. The model may not be applicable to transfer patients, who weren't included in the study population, they said. The tool is unique in that it can be readily incorporated into clinical practice via the Web-based Patient Management Tool of GWTG-Stroke, which makes an automated calculation of risk once patient data have been uploaded, they added. The risk score that included the NIHSS made even more accurate predictions, which hopefully will encourage clinicians to use the NIHSS, which is now recorded about 40% of the time, said Ralph Sacco, MD, president of the American Heart Association, in a press release.

Preprocedural statins reduce risk of postprocedure cardiovascular events

Using statins before invasive procedures significantly reduces the risk of postprocedural myocardial infarction, but only non-significantly reduces death, a recent meta-analysis found.

Researchers chose randomized, controlled trials of patients who underwent an invasive procedure and had been randomized to statin therapy or control. Control meant placebo, usual care or low-dose statin therapy. Invasive procedures were defined as percutaneous coronary intervention (PCI), coronary artery bypass grafting (CABG) or noncardiac surgical procedures, including vascular surgery. The researchers performed a literature search of the MEDLINE, Cochrane and the databases from inception to February 2010 for studies in which statins were initiated before procedures and clinical outcomes were reported. Twenty-one trials with a total of 4,805 patients met their selection criteria. The results were reported in the Sept. 28, 2010 Journal of the American College of Cardiology.

The use of preprocedural statins significantly reduced post-procedural myocardial infarction (MI) compared with control (risk ratio [RR], 0.57; 95% CI, 0.46 to 0.70; P<0.0001). A 5.8% absolute risk reduction occurred after PCI (P<0.0001) and a 4.1% risk reduction after noncardiac surgical procedures (P=0.004), but there was no statistically significant reduction after CABG. The risk reduction for PCI occurred when statin therapy was initiated approximately one to seven days before the procedure, while statins started about four weeks before noncardiac surgical procedures reduced postoperative MI. There was a nonsignificant reduction in all-cause mortality with preprocedural statin use (RR, 0.66; 95% CI 0.37 to 1.17; P=0.15). Preprocedural statins also reduced post-CABG atrial fibrillation: Among the CABG studies, postoperative atrial fibrillation was 19% in the statin arm and 37% in the control arm (RR, 0.54; 95% CI, 0.43 to 0.68; P<0.0001).

The authors concluded from the analysis that preprocedural statin therapy is beneficial, but noted that it's difficult to determine the type and dose of statins to use, or how long they should be used before a procedure. The PCI and surgical studies both used a wide variety of drugs and doses. Still, in the PCI studies, 56% of the weight of analysis came from trials with use of atorvastatin at 40 mg or more; 58% of the CABG studies' analysis entailed use of atorvastatin at 20 mg or more; and 91% of the analysis of noncardiac surgery trials entailed use of 80-mg doses of fluvastatin, they said. As for the finding that using statins preoperatively didn't reduce post-CABG MI, the authors noted the trials they found were small and involved low doses of statins, thus the issue deserves further study.

Constipation in mechanically ventilated patients associated with death, infection

Hypotension and a low PaO2/FiO2 ratio are associated with constipation in long-term ventilated patients, which in turn is associated with adverse outcomes, a recent study found.

In a prospective, observational cohort study at an academic hospital ICU, researchers studied 609 adult patients who underwent mechanical ventilation for six days or more. Patients were followed for the passage of stools and for ICU-acquired bacterial infections until ICU discharge, death, or a decision to move to palliative care. The feeding policy was to start enteral nutrition as soon as possible through a nasogastric tube with a peristaltic pump. Constipation management wasn't protocolized, but usually involved an enema with sorbitol, followed by lactulose if needed. “Early” passage of stools occurred within six days of ICU admission; “late” passage (i.e., constipation) occurred on or after six days. Results were published in the October 2010 Critical Care Medicine.

Forty-two percent of patients had early passage of stools and 58% had late passage; the two groups didn't differ significantly in baseline characteristics and had similar logistic organ dysfunction scores on the first day of mechanical ventilation. Constipation was independently associated with a Pao2/Fio2 ratio of less than 150 mm Hg (adjusted hazard ratio [HR], 1.40; 95% CI, 1.06 to 1.60; P=0.0073), systolic blood pressure between 70 and 89 mm Hg (adjusted HR, 1.48; 95% CI, 1.17 to 1.79; P=0.002) and systolic blood pressure less than 68 mm Hg (adjusted HR, 1.29; 95% CI, 1.01 to 1.60). Logistic organ dysfunction scores were higher on the fourth and ninth days of mechanical ventilation for constipated patients versus non-constipated patients, and the mortality rate for each was 30% and 18%, respectively (P<0.001). Sixty-six percent of constipated patients had acquired bacterial infections at any site, compared to 34% of non-constipated patients (P<0.001).

Neither vasopressors nor positive end-expiratory pressure (PEEP) levels were associated with constipation, which supports the notion that constipation is more related to severity of hypotension and hypoxemia rather than their treatments, the authors noted. Though an association between constipation and ICU morbidity and mortality doesn't provide definitive proof, it is possible constipation affects disease severity in ICU patients, the authors said, as constipation is a risk factor for intra-abdominal hypertension, which may cause organ dysfunction. Overall, these study results may help to identify ventilation patients at high risk for constipation, so their care can be improved, they added.

Pneumococcal urinary antigen test helps narrow CAP treatment

The pneumococcal urinary antigen test is an effective tool to guide the treatment of hospitalized adult patients with community-acquired pneumonia (CAP), a recent study found.

The prospective study included 464 Spanish patients hospitalized with 474 cases of CAP from February 2007 through January 2008. Blood cultures were used in most cases to isolate the organism responsible for the infection. For some patients, pleural fluid was analyzed and for some, paired serum samples were tested for atypical microorganisms.

Pneumococcal urinary antigen assay was performed in 383 cases and the findings were positive in 136 (35.5%). In total, Streptococcus pneumoniae was found to be the causative pathogen in 171 of the 474 total cases. The pathogen was detected exclusively by the urinary antigen test in 75 cases (43.8%). Researchers calculated the specificity of the test to be 96%, while the positive predictive value ranged from 88.8% to 96.5% and the positive likelihood ratio ranged from 14.6 to 19.9.

Physicians were able to use the urinary antigen test result to reduce the spectrum of antibiotics for 41 patients, all of whom were cured of the pneumonia. The study authors theorized that this optimization of therapy could have been possible in all 75 patients whose pathogen was definitively identified. The authors also noted that this study, unlike many others, included patients with immunosuppressive conditions (making up about 20% of the study population), so the findings are applicable to those high-risk patients as well.

A positive result on a pneumococcal urinary antigen test allows clinicians to optimize antimicrobial therapy, the authors concluded. Therefore, they suggested that the test be incorporated into clinical guidelines as a supplement, although not a replacement, for classic microbiological studies. The study was published by Archives of Internal Medicine on Jan. 24, 2011.

Antipsychotic drugs linked to VTE risk

Antipsychotic drugs may increase risk for venous thromboembolism (VTE), especially in new users, according to a recent study.

Researchers in the United Kingdom performed a nested case-control study using a national primary care database to examine the possible relationship between antipsychotic drugs and VTE risk. Patients who had a first VTE between Jan. 1, 1996 and July 1, 2007 were matched with up to four controls according to age, calendar time, practice and sex; the controls had no prior diagnosis of VTE. The main outcome measures were odds ratios for VTE associated with use of antipsychotic drugs, adjusted for comorbid conditions, and concomitant exposure to other drugs. The study results were published online Sept. 21, 2010 by BMJ.

The authors identified 25,532 eligible cases, 15,975 with deep venous thrombosis and 9,557 with pulmonary embolism, and 89,491 matched controls. Cases were more likely to have a high BMI than controls, were more likely to live in a socially disadvantaged area, and were more likely to be taking other drugs that could increase VTE risk. Among cases and controls, 0.4% had schizophrenia, 0.3% had bipolar disorder, and 1.0% had dementia, while eight cases and 31 controls had more than one of these disorders.

Risk for VTE was 32% higher in patients who had received a prescription for antipsychotic drugs in the past 24 months compared with those who hadn't (adjusted odds ratio [OR], 1.32; 95% CI, 1.23 to 1.42). At greater risk were patients who had just started taking an antipsychotic within the last three months (adjusted OR, 1.97; 95% CI, 1.66 to 2.33), those who were taking atypical versus conventional drugs (adjusted OR, 1.73; 95% CI, 1.37 to 2.17 vs. 1.28; 95% CI, 1.18 to 1.38), and those who were taking low- versus high-potency drugs (adjusted OR, 1.99; 95% CI, 1.52 to 2.62 vs. 1.28; 95% CI, 1.18 to 1.38). Absolute risks were low, the authors found, estimating that antipsychotic drugs would account for four additional VTE cases per 10,000 patients over a year of treatment in all age groups and 10 additional VTE cases per 10,000 patients in those over 65 years of age.

The study was limited by its observational design, the authors noted. For example, they were unable to determine why most patients were prescribed antipsychotics. However, they concluded that antipsychotic drugs are associated with an increased risk for VTE and that further research on this association is warranted. An accompanying editorial advised doctors to consider the low absolute risk in clinical decision making and said that the results don't support the use of antithrombotic prophylaxis in patients taking antipsychotics without any other risk factors. However, the editorialists wrote, clinicians should be alert for “the best candidates for antipsychotic treatment, such as those people with the lowest vascular risk profile who may respond to short term and low dose treatment with antipsychotics because of individual pharmacogenetic characteristics, and those who may be more susceptible to developing side effects as a result of individual vascular risk factors possibly interacting with antipsychotics.”

Hyperglycemia predicts surgical site infections

Postoperative hyperglycemia was the most significant predictor of surgical site infections identified in a recent study.

Researchers reviewed the records of more than 2,000 general and vascular surgery patients at an academic hospital. They identified a number of predictors of infection in the general surgery patients, including increasing age, emergency status, American Society of Anesthesiology physical status classes P3 to P5, operative time, more than 2 U of red blood cells transfused, preoperative glucose over 180 mg/dL, diabetes and postoperative glucose. However, once the results were adjusted for postoperative glucose level, the other variables were no longer significant predictors, indicating that postoperative glucose was the most important risk factor. There was an association between increasingly higher infection rates and glucose readings above 110 mg/dL.

A subanalysis of colorectal surgery patients found that a postoperative glucose level above 140 mg/dL was the only significant predictor of infection. However, the subanalysis of vascular surgery patients did not find postoperative glucose to be a significant predictor. The vascular patients were 1.8 times more likely to develop a surgical site infection than the general surgery patients, but operative time and diabetes were the only significant predictors identified. The results were published in the September 2010 Archives of Surgery.

The study was limited by the availability of glucose readings for only half of the patients during the first 12 hours after surgery. The authors also acknowledged that the study establishes only an association, not a cause. It may be that hyperglycemia increased the risk of infections, but it's also possible that the other risk factors led to hyperglycemia. The authors called for additional studies to confirm whether postoperative hyperglycemia is a modifiable risk factor for surgical site infections.

The study's failure to find an association between hyperglycemia and infections in vascular patients—who are more likely than average to be diabetic—is striking, according to an accompanying critique. It may have been a fault of the study design or an indication that other factors, such as A1c, tobacco use, obesity, operative time and tissue perfusion/oxygenation, are important, the critique speculated. Still, the study is important because it shows a linear relationship between postoperative glucose and infection and suggests that 140 mg/dL should be the target to minimize infection risk.

Dexamethasone as adjunct treatment helps pneumococcal meningitis patients

Dexamethasone as an adjunctive treatment for pneumococcal meningitis substantially improves the illness prognosis, a large Dutch study found.

Researchers studied 357 episodes of pneumococcal meningitis in 354 patients between 2006 and 2009. Of those, 84% were given 10 mg of dexamethasone intravenously every six hours for four days with or before their first dose of parenteral antibiotics. Results were compared to an earlier study of 352 people who were treated for bacterial meningitis in 1998-2002, before Dutch guidelines recommended using dexamethasone. In that study, only 3% of patients were given dexamethasone. In both studies, participants were assessed with the Glasgow Outcome Scale at discharge. A score of one was given for death, two for coma, three for severe disability, four for moderate disability and five for mild or no disability. Scores from one to four were considered “unfavorable,” and a score of five was “favorable.”

Characteristics on admission were comparable for both cohorts. Antimicrobial treatment consisted of penicillin or amoxicillin in 33% of episodes, third-generation cephalosporins in 28%, and a combination of penicillin or amoxicillin and third-generation cephalosporins in 34% of episodes. Another regimen was used in 5%. Adherence to antibiotic guidelines was similar between the two cohorts; guidelines haven't changed between 1998 and 2009. At discharge, an unfavorable outcome was present in 39% of patients in 2006-2009 and 50% in 1998-2002 (odds ratio [OR], 0.63; 95% CI, 0.46 to 0.86; P=0.002). Mortality rates (20% vs. 30%; P=0.001) and rates of hearing loss (12% vs. 22%; P=0.001) were lower in 2006-2009. Differences in outcome remained after adjusting for differences in case mix between cohorts via use of a multivariable prognostic model. Results were published in the October 26, 2010 Neurology.

These data are valuable because the key analysis is based on comparing two national cohorts on an intent-to-treat basis (one from a period in which hardly any dexamethasone was used and one in which dexamethasone was generally prescribed). A bias due to prescribing dexamethasone to patients who are systematically in poorer or better condition does not apply when comparing two national cohorts as a whole, the authors noted. Also, the researchers applied an extensive adjustment for differences in case mix between the cohorts based on a large amount of data on prognostic factors in bacterial meningitis, they noted. And, there weren't any substantial improvements in other types of treatment for bacterial meningitis during this time period. The decline in case fatality from 30% to 20% between the two cohorts can't be attributed to a change in disease severity as researchers corrected for many prognostic factors, they noted; thus, the main difference was the use of adjunctive dexamethasone therapy, they said.

“Hard stop” CPOE alert reduces undesired prescribing—at a cost

An alert on computerized physician order entry (CPOE) systems that warned against ordering conflicting medications helped lower undesired prescribing, but had unintended negative consequences for several patients, a study found.

Researchers randomly assigned 1,981 clinicians to an intervention group that used a CPOE with an alert to reduce concomitant orders for warfarin and trimethoprim-sulfamethoxazole, or a “usual practice” control group. The “hard-stop” alert appeared as a pop-up window notifying a clinician that the order couldn't be processed because of a significant potential drug-drug interaction. It was triggered whenever a clinician placed an order for trimethoprim-sulfamethoxazole with an already active warfarin order, when warfarin was ordered for a patient already taking trimethoprim-sulfamethoxazole, or when both were ordered at the same time. An alert could be overridden by calling a pharmacist or entering an indication of Pneumocystis carinii pneumonia prophylaxis. The control group used the usual practice of a pharmacist calling prescribers to notify them of an interaction and recommend cessation of concurrent orders. The study involved 1,872 residents and 99 nurse practitioners at two academic medical centers in Philadelphia between August 2006 and February 2007. Results were published Sept. 27, 2010 in Archives of Internal Medicine.

There were 194 unique alerts in the intervention group and 148 in the control group. Of the alerts in the intervention group, clinicians had the desired response (not reordering the drug) 57.2% of the time, compared to 13.5% of the time in the control group (adjusted odds ratio, 0.12; 95% CI, 0.045 to 0.33). In both cases, the undesired response most often involved ordering warfarin. The greatest proportion of desired responses was observed in the first three months of the intervention. The institutional review board stopped the studya month early because of four instances of unintended consequences in the intervention group—delayed treatment with trimethoprim-sulfamethoxazole or delayed warfarin initiation when these drugs were warranted. Electronic records review showed that no infections or thrombotic complications could have been related to these delays, however, the authors said.

The decreased effectiveness of the intervention over time may be due to the fact that the clinicians usually worked in teams, and may have worked alongside someone in a different study group (though researchers tried to prevent this), thus spreading awareness of the interaction alert, the authors said. Researchers didn't review the medical records or outcomes of all patients who went on to receive concurrent medications despite alerts, so they were unable to conclude whether the benefit of reducing the drug interaction may have outweighed the harm that led to the trial's early termination, they said.

Given that the U.S. will soon invest $50 billion in health information technology and CPOE is a key part of that, it's important to nail down the best use of CPOE, the authors and a commentator noted. For one, there needs to be more discrimination with alerts so that physicians don't get alert fatigue. “Clinicians tend to override alerts because they are perceived to be nonspecific and lack the clinicians' additional knowledge of the clinical situation for the specific patient context,” the authors wrote. The strength of a CPOE alert should be related to the severity and importance of a drug-drug interaction, the commentator said, with hard stops used only when there are no exceptions.